Vista's Limited Symlinks

An anonymous reader writes, "Symlinks haven't really been added to Windows Vista. It seems that the calls to the Windows Vista symlink API only occur during the creation of such files or when accessing them from Windows Explorer. What this means is, you can't access symlinks from another OS. To be fair, you probably didn't expect to be able to dual-boot into XP and suddenly have access to the symlinks you created on the Vista partition earlier that day. But then again, you probably expected to be able to access these symlinks through a network share/UNC path or as files on a webserver. But you can't." From the article: "Clearly, Vista's symlink API isn't complete — hopefully this is something that can be patched via a hotfix and that we don't have to wait for Fiji to get something as simple as UNC support built in."

Comments

Broken Window syndrome

[davidwr]

Time to call in the code inspectors.

Shorter titles are Sweeter

[Hexstream]

Vista's limited.

Shortcuts are nothing new

[Electrode]

This sounds like the "shortcut" feature that's been around since Win95. Have they actually implemented something new here, or just given some old junk a new name?

Re:Shortcuts are nothing new

[kimvette]

No, what is being discussed here is links, e.g., creating an additional filename referencing an inode.

http://win32.mvps.org/ntfs/lnw.html
http://en.wikipedia.org/wiki/NTFS_symbolic_link
http://answers.google.com/answers/threadview?id=34 1355

NTFS does support links, but as usual from Microsoft, it's half-baked and only the bare minimum required for POSIX compliance was implemented. From sysinternals (now a Microsoft site) you can download a utility for manipulating NTFS links, or you can install the free Services for Unix (again, from Microsoft's web site) to get the M$ version of ln.

Re:Shortcuts are nothing new

[wesnerm]

Your information about MS half-baked implementation of links only applies to Windows XP. Vista introduced true UNIX-style symbolic links in addition to the junctions introduced in earlier versions of NTFS. The new links were incorporated from Windows Services for UNIX. See my posts on Vista Symbolic Links from a month ago. http://wesnerm.blogs.com/net_undocumented/2006/10/ symbolic_links_.html http://wesnerm.blogs.com/net_undocumented/2006/10/ symbolic_links__1.html

Re:Shortcuts are nothing new

[Knuckles]

Vista introduced true UNIX-style symbolic links

The article is about how it doesn't.

Re:Shortcuts are nothing new

[Duggeek]

Shortcuts are ill-conceived "placecards" for files and executables; like a road-sign, they can become outdated by changes to the target file and don't necessarily reflect any relevant properties of the target. The Windows95 implementation marks the change from a centralized database of "links" on the system to independent, "shortcut files". Neither implementation was directly linked to the filesystem.

Symlinks are directly tied through the file-system (not through a secondary API); in my analogy, a symlink is more like GPS navigation, it takes you all the way to the target.

Best-in-show would go to MacOS for aliases; more like a portal to the target, and if the target moves, the alias corrects itself.

In effect, the Windows-style "shortcut" is a complete dupe of *nix symlinks, (which came first) only bastardized and "extended" by M$ engineers. The fact that MS-DOS (ergo, Windows) used FAT file system negated the possibility of using symlinks. The data-structure was so simplified, there was no room left for advanced tagging or linking mechanisms. (already present in 'ext' at that time)

It's been years since MS bought several (disputed) Unix patents from SCO, and it's apparent they're still not making anything useful of those. Another case of wrecking already-working code for the sake of thumbing their noses and saying "It's original code! You gotta pay us for it now!"

Any wonder why they named it "shortcut"?

Re:Shortcuts are nothing new

[Anonymous Coward]

No, what is being discussed here is links, e.g., creating an additional filename referencing an inode.

That would be a "hard link". Grandparent was right, sym(bolic) links (unix "ln -s") are like shortcuts: rm the symlink and you don't erase the file it was linking too (whereas you would if the link was referencing the same inode). The only difference is that Windows Vista now implements symlinks directly in the filesystem instead of through .lnk files.

Re:Shortcuts are nothing new

[emurphy42]

Symbolic links can also become outdated if the target moves. It's hard links that can't.

Re:Shortcuts are nothing new

[rm69990]

Microsoft didn't buy the "patents", they "licensed" them. The only funny thing is that SCO owned (and maybe still does) only one patent, and it has nothing to do with operating systems :-P Neither SCO nor Microsoft ever claimed that Microsoft actually purchased patents, nor does the Wikipedia article you linked to.

The license covered Unix code as well as far as I can tell, most of which is in the public domain, and the remaining bits owned mainly by Novell and other parties.

Re:Shortcuts are nothing new

[listen]

If you rm a one link to an inode, (or close a file descriptor), it reduces the inodes usage count by one. When the count reaches zero, the blocks referenced by the inode are marked free. The behaviour you describe would be utterly horrendous.

Re:Shortcuts are nothing new

[Ucklak]

It also sounds like a puppy granting wishes but that isn't what the article is about.

That so called "shortcut" feature that has been around since Windows 3.1 isn't a 'link' as it works on POSIX complaint systems. It is a shortcut that gives the user - not the computer - a quick way to access a directory. It cannot handle I/O functions.

Re:Shortcuts are nothing new

[Anonymous Coward]

Hmmm. How odd. Wesnerm has only ever posted one comment (the one above), which would lead me to believe that MS is beginning to ramp up their astroturfing campaign. At the same time, he has a UID in the 600000s, which means he's been kicking around here for a few years.

Re:Shortcuts are nothing new

[Tim Browse]

I'm slightly confused - hard links have been in NTFS for years, and I've use them a few times for various tasks, so why is NT so bad in relation to others? I'm also not sure why shortcuts are different to symlinks. Don't both just store a pathname to link to? If the target changes, the symlink/shortcut still points to that named file, or if the target no longer exists, both shortcuts and symlinks no longer work? Or am I misunderstanding? Your analogy doesn't seem to work - the key difference as far as I can see is that symlinks behave as if they were files in every way - shortcuts are just signposts for Windows Explorer. Assuming you're using Explorer, they seem to work the same.

As for Mac OS X aliases being 'best of class', the only time I needed to use one of those I found it didn't work over SMB sharing - exactly the same sort of complaint as is being levelled at Vista, as far as I can tell. This was a few revs back of OS X, so maybe it's fixed now.

Re:Shortcuts are nothing new

[spitzak]

Hardlinks are what Unix had *originally*, so they are even older. Symlinks were added later, as in fact hard links were very hard to manage. Permanent hard links are almost never used on Unix today (they are used for short-term things such as renaming files atomically).

There are "semi hard" links that Mac OS (but not OS/X?) do that may be better than both symbolic and hard links. They act like hard links as long as the linked-to file exists. If it is deleted they then act like a soft link (not sure if they revert to the original name or to some name computed from the most recent location). If a file is created at the soft link location they then go back to acting like a hard link.

Oddly enough, even though people pretty much admit that hardlinks are not as useful as symbolic links, Microsoft has done *more* to emulate hard links than symbolic ones. They are actually copying one of the more useless and harder to emulate Unix features, and not what is needed, which are symbolic links which are trivial to emulate. They are of course desperately afraid of making it too easy to emulate a Unix environment on Windows, and will do anything, including make up bogus excuses about the difficulty, to avoid it. This is also why Notepad still craps out if there are bare LF in the file and inserts a "utf-8 bom" at the start to break any Unix utilities that look at the first bytes.

Re:Shortcuts are nothing new

[WhoBeDaPlaya]

Can't believe people are still actually using that PoS program that MS calls notepad. Been using Metapad for ages : http://liquidninja.com/metapad/

Re:Shortcuts are nothing new

[profplump]

Aliases should work via SMB, so long as you've still got the resource fork and are accessing it from another Mac. They don't work from Windows for obvious reasons -- Windows has never implemented a volume-ID to path resolution required systems, as is required by the alias structure.

I'll give you that this makes aliases less flexible, but aliases give you the ability to make a link to a file on a disk image on a remote disk -- and rename any of the files, disks, or folders involved -- while still being able to double-click to open. All of the necessary mounting and name correction just happens; it's like a hard-link but without the need for the file to be on the same disk (or even local).

Re:Shortcuts are nothing new

[Anonymous Coward]

> NTFS does support links, but as usual from Microsoft, it's half-baked and only the bare minimum required for POSIX compliance was implemented.

NTFS's links have nothing to do with POSIX compliance, and were implemented for DFS.

Re:Shortcuts are nothing new

[nuzak]

> In effect, the Windows-style "shortcut" is a complete dupe of *nix symlinks, (which came first) only bastardized and "extended" by M$
engineers.

Oh wow, you spelled MS with a dollar sign. Aren't you clever. Were you even born long enough ago to know what a .PIF file is?

Oh but hey Unix invented filesystems, right?

Re:Shortcuts are nothing new

[lgw]

Hardlinks AND symlinks have been in NTFS forever, but because you can't create them through the GUI or the DOS prompt, most people don't knwo they exist. Shortcuts are very different from symlinks, as shortcuts are just a GUI thing. If a program is hard-coded to look for "C:\foo\bar.ini" you can't move bar.ini elsewhere and replace it with a shortcut, as bar.ini.lnk isn't helpful. You can replace it with a link, however.

To create a symlink (or hardlink) in NTFS you either need to use Unix services for Windows to get a command line, or use the POSIX compatibility engine programatically (just make the same API calls you would in UNIX though random function names will have leading '_'s just to make it difficult).

Re:Shortcuts are nothing new

[kimvette]

Please tell us how you cd to a shortcut, then we can all agree with you.

A shortcut is analagous to a KDE .desktop file, not to a filesystem symbolic link.

Try again.

Re:Shortcuts are nothing new

[JohnVanVliet]

Why would anyone WANT M$'s version of " ln " MinGW ln is just fine in windows

Re:Shortcuts are nothing new

[TheRaven64]

MacOS Aliases are close to symlinks, but they point to a unique file hash, rather than a file name. This means you can move the target around (even between mounted volumes) and the link will still work as long as the file is still on a mounted volume (the file hashes are indexed by HFS[+], so they are quick to look up). Old MacOS applications were encouraged to use the file hashes instead of the file name for things like their 'recent documents' menu, since it would allow documents to be accessed even if they were moved.

OS X still supports them, and they are exposed through the /.vol hierarchy (or, more sensibly, the Carbon APIs), but their use is now discouraged.

Re:Shortcuts are nothing new

[TheRaven64]

Shortcuts and symlinks are almost the same thing; they are both files that store the path of the 'real' file. The difference is how they are implemented. Symbolic links are implemented in the VFS layer; if you do a standard system call to operate on a file, and pass it the path to a symlink, it will operate on the target. Extra APIs are provided for manipulating symlinks directly. Shortcuts, on the other hand, are just regular files. An application has to understand that a .lnk file is a shortcut, and use the appropriate APIs to access the target file. In short, UNIX applications that pre-date symlinks could work with symlinks transparently, while Windows applications that predate shortcuts couldn't work with them.

It's slightly odd that it should be this way around, since Microsoft typically does everything it can to make developers lives easy.

Re:Shortcuts are nothing new

[spitzak]

If you double-click a text file on a standard Windows install you will get Notepad. Otherwise the problems with Notepad would not matter.

I am quite convinced Microsoft does this on purpose so that files imported from Unix will look like crap, and if you save the file it will break on Unix. Wordpad, Word, the IDEs, and almost any other program they write does not have this behavior, but the default one you get for text files does. What other explanation is there?

Re:Shortcuts are nothing new

[porl]

on this note, is there any good non-geek article i can use to show people the difference between the two? i must not be that great at explaining some things (i know the difference myself, but i mean a simple analogy type article thing.. my employers love them). :)

Re:Shortcuts are nothing new

[Mr2001]

Wow, talk about paranoid. Notepad exhibits that behavior because it's just a thin wrapper around the standard text box control. Remember when you couldn't load any files bigger than 64k in Notepad? That was because of an OS limit too. The standard text box isn't meant to be a full-featured text editor. Notepad's flaws are the result of laziness, not malice.

Here's another nail in the coffin of your theory: how many people could there possible be who import text files from Unix systems--or who even know what Unix is--but still aren't sophisticated enough to figure out how to open them with Wordpad (or good old EDIT.COM), which has no problem with Unix line breaks?

Re:Shortcuts are nothing new

[Tim Browse]

Aliases should work via SMB, so long as you've still got the resource fork and are accessing it from another Mac. They don't work from Windows for obvious reasons -- Windows has never implemented a volume-ID to path resolution required systems, as is required by the alias structure.

I was using it from a Windows machine, but I was confused (and still am after your explanation), as to why aliases didn't work over SMB. You say it's 'obvious'. I thought an alias was a file system level object, from what I remembered when I read about them in 'Inside Macintosh'.

So, to me, it's irrelevant that Windows has never implemented a 'volume-ID to path resolution' system, as you're doing network sharing, so you already have a virtualisation layer present. If I use Windows to open a folder shared on a Mac via SMB, and that folder has an alias to another folder on the Mac, why doesn't that just appear as a standard folder via SMB? Isn't that the point of aliases? They 'just work', because they're low enough in the FS system to just work as files or folders? Obviously, if you want to, you can detect aliases (so, e.g. you don't back up the same folder twice in a backup program), but claiming it's the fault of the client machine or the network protocol seems a bit lame.

SMB supports files and folders, so aliases should 'just work', even over SMB. The fact that they don't seems to me to be the same failing as Vista is being berated for in this story.

Re:Shortcuts are nothing new

[Tim Browse]

Hardlinks AND symlinks have been in NTFS forever, but because you can't create them through the GUI or the DOS prompt, most people don't knwo they exist. Shortcuts are very different from symlinks, as shortcuts are just a GUI thing. If a program is hard-coded to look for "C:\foo\bar.ini" you can't move bar.ini elsewhere and replace it with a shortcut, as bar.ini.lnk isn't helpful. You can replace it with a link, however.

Ah, thanks for clearing that up - I thought that NTFS supported symlinks for ages, too, but the article and comments seemed to suggest it was new, and I was only sure that I'd use hard-links before (I just download one of the hardlink command-line utilities/explorer context menus whenever I need to use links), so I assumed I was wrong about symlinks.

And yeah, shortcuts are quite sucky due to the .lnk thing.

Re:Shortcuts are nothing new

[SanityInAnarchy]

You don't understand the difference between hardlinks and symlinks. Try again.

A hardlink has the behavior you describe. A symlink is a file which refers to another file by name, not by inode. You can have multiple hardlinks to the same symlink, as the symlink itself is just a file, and a separate file from the one it links to.

Re:Shortcuts are nothing new

[Yamhill]

Not quite non-geek, but an amusing read even if you're all over hardlinks, symlinks, etc.: http://shell-shocked.org/article.php?id=284/

Re:Shortcuts are nothing new

[Anonymous Coward]

The article is saying that a vista smb server won't magically translate symlinks for an xp client. They don't understand that that's not how symlinks work. The client side interprets the symlink. nfs clients interpret symlinks from nfs servers. automount systems work because the client is interpreting the symlink and mounting the new filesystem transparently. thinking that the server should translate symlinks shows a complete lack of understanding of the technology.

Re:Shortcuts are nothing new

[profplump]

First, if an alias was purely a file-system-level object, I don't see why you think it would work across file systems. The native file system of Apple OSes is HFS or HFS+; SMB is another animal altogether. Even if HFS implemented aliases in the manner you imagine, it still wouldn't work via SMB, just as symlinks don't work on SMB, since SMB doesn't provide support for either mechanism. If it was a native file-system object it would limited by the scope and functions of the file system on which it was contained.

That's not to say aliases aren't tied into the file system -- they contain information native to whatever file system(s) they reference, and in that respect are part of the file system. But their scope is wider, so they cannot be considered native objects, as an single file system may not contain sufficient information to resolve them. They are however, a native part of the VFS on Apple OSes, so from the perspective of anyone using an Apple OS, they are a native object, just as symlinks are part of the linux VFS and therefore can be used even on file systems that do not natively support such an object.

Aliases do "just work" over SMB, so long as the system mounting the SMB file system supports aliases. You can make aliases to files and folders on an SMB share, or store aliases on an SMB share, and any system that knows how to read aliases will happily resolve them for you. But aliases are not a native part of SMB sharing, so simply having an SMB client implementation is not sufficient to resolve aliases.

Aliases are not superior in all cases to symlinks or hard links, but they are very useful for users creating links to files or folders, particularly if you may not have all the appropriate paths mounted, as is often true with remote disks or disk images. They are similar to the new system Vista is providing, but they are resolved at runtime in a manner that can transparently resolve changes in file paths and disk locations and names.

Re:Shortcuts are nothing new

[Duggeek]

True, it was a licensing purchase. Another directly related article does point out SCO's claim. I'm sure many would appreciate it if you would cite your sources when making such specific claims, I know I would.

The summary of Linux controversies clearly shows that M$ put an interest into SCO's so-called "patent rights" in mid-2003... though the scope of the licensing is only a fraction of what Novell acquired from AT&T. (and still holds today) Certainly not enough for SCO to back-up their own notion that they "own" Unix.

Regardless of Novell's current Linux offerings, the fact that they have a succinct hold on the origins of Unix has a yet-to-be-determined impact on the future of licensed Unix platforms.

Add to that, they've got into bed with M$, and there's but a hint of what was actually purchased with the $350 mega-bills. Though Microsoft has their "Q&A", Novell is apparently all aglow about the new agreement.

This turns out to be off-topic, I know. I truly want to leave FUD aside with all this, however the details are speaking more loudly than my keystrokes could ever accomplish. This issue is already part of Novell's wiki entry and even throws some props to us /.'ers.

Besides, symlinks are damn useful. Unlike shortcuts, they have a distinct file attribute, but do not behave like their target (an advantage, to be sure); shortcuts are nothing but special files, and can be surreptitiously re-assigned to a new target, even a URL. Just check out these vulnerabilities regarding their use. I haven't seen any such problems with symlinks.

Re:Shortcuts are nothing new

[rm69990]

To back up which claims? That SCO only holds one patent?

Simply search for them at uspto.gov (the site doesn't let you link directly to search results or patents for some odd reason). The only patent currently held by SCO is:

  Method and apparatus for executing multiple JAVA(.TM.) applications on a single JAVA(.TM.) virtual machine Patent # 6,931,544

As for the claim that SCO licensed code to Microsoft and not patents, I refer to SCO's own SEC filings, and what only they can only modify and edit as opposed to Wikipedia's policy of allowing any random tom, dick and harry to edit entries.

From the filings:

We initiated the SCOsource effort to review the status of these licensing and sublicensing agreements and to identify others in the industry that may be currently using our intellectual property without obtaining the necessary licenses. This effort resulted in the execution of two license agreements during the April 30, 2003 quarter. The first of these licenses was with a long-time licensee of the UNIX source code which is a major participant in the UNIX industry and was a "clean-up" license to cover items that were outside the scope of the initial license. The second license was to Microsoft Corporation ("Microsoft"), and covers Microsoft's UNIX compatibility products, subject to certain specified limitations. These license agreements will be typical of those we expect to enter into with developers, manufacturers, and distributors of operating systems in that they are non-exclusive, perpetual, royalty-free, paid up licenses to utilize the UNIX source code, including the right to sublicense that code.

The amount that we receive from any such licensee will generally depend on the license rights that the licensee previously held and the amount and level of our intellectual property the licensee desires to license. The two licensing agreements signed by us to date resulted in revenue of $8,250,000 during the April 30, 2003 quarter and provide for an aggregate of an additional $5,000,000 to be paid to us over the next three quarters. These contracts do not provide for any payments beyond 2003, except that Microsoft was granted the option to acquire expanded licensing rights, at its election, that would result in additional payments to us if exercised. In connection with the execution of the first license agreement, we granted a warrant to the licensee to purchase up to 210,000 shares of our common stock, for a period of five years, at a price of $1.83 per share. This warrant has been valued, using the Black-Scholes valuation method, at $500,000. Because the warrant was issued for no consideration, $500,000 of the license proceeds have been recorded as warrant outstanding and the license revenue reduced accordingly.

http://sec.gov/Archives/edgar/data/1102542/0001104 65903012299/j2045_10q.htm

Nowhere in the filing does it mention any patents being licensed whatsoever to anyone. Once Novell started demanding royalties for these licenses, SCO told everyone it was a patent license, despite their SEC filings saying otherwise.

http://www.novell.com/licensing/indemnity/pdf/6_24 _03_n-sco.pdf

It happens quite often that SCO's SEC filings conflict with their public statements and court filings.

Novell has recently filed for an injunction against SCO, stating unequivocally that those agreements were source code licenses and thus Novell gets 95% of the revenue. Novell states that the license is a copyright license.

Re:Shortcuts are nothing new

[Duggeek]

Unix invented filesystems, right?

You bet your momma's table-sized hard-disks. While there may be contentions of technical accuracy, I postulate that Bell Labs Unix provided the groundwork for just about all enterprise filesystems in use today. (also about the time I was born... so yes, I do remember)

There's a distinction to be made here... a PIF file is not a "link" in this context. It can however be called a "launcher"... since it specifically contains data relevant to the environment for an executable.

While it does have a way to "Find Target", (Win95+) it ultimately gets left behind (broken) unless updated by installer applications or the user. Read carefully and you'll see that it's even older than LNK files; all the way back to MS-DOS.

Linux also distinguishes between links and launchers, however the differences don't give much potential for abuse. (unlike LNK/PIF files) The only exception is the symlink race vulnerability, but is a marginal risk at best with current security measures.

I'm sure that nobody forgot the beloved PIF, but who really considers them to be "links" anyway?

Re:Shortcuts are nothing new

[Knuckles]

"Interprets", yes:

When an NFS client does a stat( ) of a directory entry and finds it is a symbolic link, it issues an RPC call to read the link (on the server) and determine where the link points. This is the equivalent of doing a local readlink( ) system call to examine the contents of a symbolic link. The server returns a pathname that is interpreted on the client, not on the server.

But the symlink is a feature of the on-disk filesystem and is there whether the client does something with it or not.

Re:Shortcuts are nothing new

[Hal_Porter]

To create a symlink (or hardlink) in NTFS you either need to use Unix services for Windows to get a command line, or use the POSIX compatibility engine programatically (just make the same API calls you would in UNIX though random function names will have leading '_'s just to make it difficult).

You need to

#define _CRT_NONSTDC_NO_DEPRECATE 1

or add a /D_CRT_NONSTDC_NO_DEPRECATE in the compiler options.

http://forums.microsoft.com/MSDN/ShowPost.aspx?Pos tID=87401&SiteID=1

Re:Shortcuts are nothing new

[Hal_Porter]

It's because shortcuts are from the Windows 95 shell, which had to work on FAT. Shortcuts can link to things that aren't files too, IIRC. So they're not really anything to do with symlinks.

And they're only used in the Start Menu, so only the shell supports them.

Someone reverse engineered the .lnk file format

http://mediasrv.ns.ac.yu/extra/fileformat/windows/ lnk/shortcut.pdf

NTFS supports real hardlinks, but it's only used by POSIX. You can use DeviceIoControls to make them from Win32, and sysinternals has sample code.

Oddly enough in FAT32, the high 4 bits of each FAT entry are reserved. The problem with a FAT based filesystem is that you don't have an inode to hold a count of the number of links to a file. But you could use the high bits of the fat entry to track it. It'd be kind of gross though, presumably files with more than 16 links would need to add extra clusters to hold the extra counter bits. E.g. a one cluster file would have 4 bits of link count, a two cluster file would have 8 bits and so on. Would be hairy trying to update it though, since you'd need to have some kind of mutex to protect the FAT.

Re:Shortcuts are nothing new

[spitzak]

how many people could there possible be who import text files from Unix systems

Answer: EVERY SINGLE F**KING ONE OF THE WINDOWS USERS HERE. It may be an unknown mystery to you but we have something called a network here and our Windows machines are able to see and read/write the contents of our Unix disks.

I'd like to know how they managed to add utf-8 support to Notepad, remove that 64K barrier you mentioned, and still not fix the linefeed problem. And why Word and wordpad and VC++ IDE and every single other program I try on Windows has no problem handling bare linefeeds (they may write cr+lf, but I consider failure to read that a Unix problem that should be fixed there).

Also the "standard textbox" displays linefeeds as newlines. It has to, considering you can give it a constant string from in-memory in a C program. What you are talking about is a specialized widget that edits file contents only.

Re:Shortcuts are nothing new

[Mr2001]

[quoted: how many people could there possible be who import text files from Unix systems]

Answer: EVERY SINGLE F**KING ONE OF THE WINDOWS USERS HERE.

Whoa, settle down there, cowboy! That sentence wasn't done. Continue reading it and you'll discover there are more words at the end which change its meaning!

Yes, lots of us slashdotters use Unix text files. But we're also clever enough to realize that Windows comes with more than one text editor, so we can open them with Wordpad or edit.com if we just need to read them, or download a decent editor if we need to write them too. So I repeat, how many Windows users do you think there are who need to work with Unix text files but are too stupid to figure out how to do it?

I'd like to know how they managed to add utf-8 support to Notepad, remove that 64K barrier you mentioned, and still not fix the linefeed problem.

They didn't add it to Notepad, they added it to the multiline edit control - Unicode support and the removal of the 64k barrier were some of the improvements made in NT. Linefeeds weren't a priority, presumably because the multiline edit control isn't supposed to be a complete text editor, and it's not difficult for apps to convert line breaks themselves.

And why Word and wordpad and VC++ IDE and every single other program I try on Windows has no problem handling bare linefeeds

Easy: they don't use the standard multiline edit control, they use the rich edit control or their own custom controls.

Also the "standard textbox" displays linefeeds as newlines. It has to, considering you can give it a constant string from in-memory in a C program.

I think you're mistaken. According to MSDN and the MS developer blogs, you need to use \r\n for a line break, not just \n. See, for one example, the comments here.

What you are talking about is a specialized widget that edits file contents only.

Something tells me you aren't a Windows developer.

Re:Shortcuts are nothing new

[Dahan]

You don't understand the difference between hardlinks and symlinks. Try again.

And you need to work on your reading comprehension. The AC incorrectly said, "rm the symlink and you don't erase the file it was linking too [sic] (whereas you would if the link was referencing the same inode)." It is not the case the rm-ing a hardlink will erase the file it was linking to. And what's with the qualification of "if the link was referencing the same inode"? Hard links reference the same inode by definition. listen was correcting AC's misconception of hard links; I don't know how you interpreted the post as having anything to do with symlinks. As you admit yourself, listen is correct: "A hardlink has the behavior you describe."

Re:Shortcuts are nothing new

[emurphy42]

To be fair, his people might be typical Windows end users who don't particularly know nor care that some of their mapped network drives happen to have Linux behind them.

Re:Shortcuts are nothing new

[emurphy42]

My first full-time employer (95-96) used hard links as follows:

We had a base package (all programs set no-write/delete), and some clients with a handful of custom mods apiece. So we had a script that created a testing directory for a new client (initially populated with hard links to the base package's programs), and another script that removed a specified hard link and replaced it with a separate writeable copy.

Bingo, each testing directory's disk usage is limited to data plus modified programs, you can tell what's been modified in it by looking at which programs aren't hard links, and the base package can't be hosed except by root.

At my current job, if we need a testing directory at all (a lot of my work is just writing custom reports which is safe to do in production), we tend to keep them on the client's network which we can remote-access at need.

I have yet to work anywhere that uses multiple-programmer teams routinely enough to use formal CVS, though it might be argued that one or two of my employers should have at least given it a hard look.

Re:Shortcuts are nothing new

[spitzak]

Seems like that could be done with symbolic links as well, though.

Re:Shortcuts are nothing new

[emurphy42]

I suppose it could have been, but are there any advantages? I can think of at least one disadvantage ('ls -l' gets ugly - remember, this couldn't be done at a folder level, due to the need to selectively fork a few files).

Re:Shortcuts are nothing new

[nuzak]

> I postulate that Bell Labs Unix provided the groundwork for just about all enterprise filesystems in use today.

Symlinks might possibly be a Unix invention, but I've read plenty of docs that predate Unix that refer to files and directories. I'm under the impression that Unix simply whittled down the Multics filesystem -- some would say to an elegant core, some would disagree (I sort of sit the fence).

> There's a distinction to be made here... a PIF file is not a "link" in this context.

Precisely. I was trying to say in a very snarky way the same point -- that they're really just a small evolution on .PIF files. A .LNK is pretty much identical in concept to a .PIF, and doesn't really try to be a symlink (cygwin does a pretty good job treating them as such, but cygwin is rather a hack, and its own category). I apologize for the tone -- hopefully last week's vacation has mellowed me out some ... I kind of go in cycles :)

obligatory quote

[laurent420]

"Those who do not understand UNIX are condemned to reinvent it, poorly." -Henry Spencer

Re:obligatory quote

[sco08y]

"Those who do not understand UNIX are condemned to reinvent it, poorly." -Henry Spencer

And those who understand Unix wrote the Unix Hater's Handbook...

Re:obligatory quote

[finity]

If you reinvent Unix poorly, does that make it a decent OS?



Laugh.

Re:obligatory quote

[rolfwind]

And those who understand Unix moved onto Plan9.

Re:obligatory quote

[drgonzo59]

No. Those who understand Unix use ... Plan9! /ducks

Re:obligatory quote

[shmlco]

"Those who do not understand UNIX are condemned to reinvent it, poorly." -Henry Spencer

Isn't that the definition of Linux? Or was that, "Those who can not afford UNIX..."

(ducks and runs)

Re:obligatory quote

[SQL Error]

And those who understand Unix wrote the Unix Hater's Handbook...

No, they're just Lisp users. They're like humans, only their brains are inside out.

Re:obligatory quote

[DrSkwid]

You're right there!

symlinks suck

per process namespaces rule

go glenda

Re:obligatory quote

[SanityInAnarchy]

Actually, it makes NT. Apparently not, then.

Re:obligatory quote

[Anonymous Coward]

No, I guess it makes for a deScent OS...
Sorry, I couldn't resist :-$

Neosmart's limited servers

[BadAnalogyGuy]

PHP has encountered an Access Violation at 7C8224B2

Re:Neosmart's limited servers

[TheShadowzero]

Maybe they tried to symlink to the article...

Re:Neosmart's limited servers

[Computer Guru]

I'm the administrator @ NeoSmart Technologies.... It's not a IIS failure, it's a PHP bug. View all links related to this particluar bug in PHP 5.1.6 The site will be back up soon! http://www.vbulletin.com/forum/showthread.php?t=20 4522 Damn PHP......

Re:Neosmart's limited servers

[Computer Guru]

It's back up now http://neosmart.net/blog/archives/285 . Enjoy!!

Re:Neosmart's limited servers

[Hal_Porter]

It's a pity the address made up of ASCII bytes.I need some more web server space pretty badly.

Access violation at 0x41414141 always makes me smile when I've been sending user names like "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA"

You just have to buy the upgrade..

[Anonymous Coward]

Let's get real. If the OS was perfect nobody would buy the next version. I bet that 80% of purchases are made by people that secretly hope there's finally a version of Windows that just works.

[and there's of course the not-invented-here syndrome - maybe symlinks are GPL-ed? :-)]

Re:You just have to buy the upgrade..

[Jugalator]

I think you're statistics is completely wrong. That 80% would probably rather be "by people who buy a new computer".
Also you need to define "just works" -- you seem to mean feature set perfection, competing operating systems don't have that either, either in features or lack of bugs.

Re:You just have to buy the upgrade..

[CoolCat23]

Well, the truth is, 95% buy it : (1) with a new computer, and (2) to play games that game developers didn't bother to make win2k or winXP compatible... Dont forget that the #1 reason people (not companies) upgrade their copy of Windows is GAMES. Have 50% of the new games run natively on Linux, and Windows market share in homes will decrease by 50% in years.

Re:You just have to buy the upgrade..

[tgcid]

Have 50% of the new games run natively on Linux, and Windows market share in homes will decrease by 50% in years.

Done.

Nope, Windows marketshare didn't change. Maybe if 50% of the new games only ran Linux.

Re:You just have to buy the upgrade..

[a.d.trick]

If the OS was perfect nobody would buy the next version

While that is a common theme with proprietary software development, I don't think that's what's happening here. I doubt that MS has any plan to implement soft-links in the nearby future. To do so would be a great help for several cross-platform applications, and Microsoft does not like cross-platform anything.

Re:You just have to buy the upgrade..

[It'sYerMam]

How about 50% of PC games [i]that you buy.[/i] It's pretty obvious.

Re:You just have to buy the upgrade..

[dbIII]

Let's get real. If the OS was perfect nobody would buy the next version.

What is normally supposed to happen is even when the bugs are all fixed in the current version people go out and buy the new version for the new feaures.

MS have the situation where they give you features in updates as well as bug fixes - which really gives no compelling reason to upgrade from Win2k in a lot of cases.

Re:You just have to buy the upgrade..

[Anonymous Coward]

As for me, "just works" means you don't have to be an expert and be ultra-careful about what you do in you system just to keep the Microsoft "ecosystem" from collapsing over its own weight every 6 months.

"Links" are not necessarily a fs feature

[Etcetera]

... not when the ability to successfully resolve something may depend on things handled by the directory/metadata structure, with tagging that is indeed OS-specific.

MacOS has had aliases since System 7 and they're far more useful than a unix-style symlink ever has been for me -- in part because everything that needs to open a file on the Mac uses the MacOS APIs. POSIX is "closer to the metal" and therefore pays a price in lost features of abstraction.

If you want Unix-style utilities to work with the new Vista symlinks, then patch the damn copy of "ls" yourself. And then submit it to the FSF.

Re:"Links" are not necessarily a fs feature

[TheShadowzero]

If you want Unix-style utilities to work with the new Vista symlinks, then patch the damn copy of "ls" yourself.

Try that again. The command to create symlinks is ln (link), not ls (list).

Re:"Links" are not necessarily a fs feature

[MORB]

POSIX isn't supposed to be closer to the metal than the native OS api. It's supposed to be a standard, non os-specific way to access files.

It's the job of the OS makers to provide a POSIX implementation that works on top on the native api. You'll find that developers writing cross-platform apps are not fond of having to conditionally use platform specific code, especially in situations like file io that has been standardized for ages.

Apparently...

[Anonymous Coward]

You can't squirt files across OS's yet.

Yeah, sure.

[dangitman]

Link, please.

Re:Yeah, sure.

[cciRRus]

It will appear when you use Windows Explorer. ;)

To be frank?

[thombone69]

A whole LOT of things about Vista are incomplete! (Like, perhaps, Vista itself?)

Re:To be frank?

[Scarletdown]

A whole LOT of things about Vista are incomplete! (Like, perhaps, Vista itself?)

So this means that Windows still isn't ready for the desktop? Perhaps the next version after Vista then...

Re:To be frank?

[Jugalator]

Reality check: I don't hear large crowds of desktop users ask for proper symlink support.
In that case, proper Windows software (including game) compatibility is more important.

Re:To be frank?

[Anonymous Coward]

>So this means that Windows still isn't ready for the desktop? Perhaps the next version after Vista then...

Ah, the sound of fanboys whining because some shitty OS is crushing their even shittier OS in the marketplace. I love making fun of you losers.

Same ol' show

[marcello_dl]

A couple of days ago, the ranting of some MS manager about interoperability, here on slashdot. But when it's time to ship, having working symlinks (rocket science apparently) for basic interoperation purpose is not there. Same old Microsoft, expect same old frustrations with Vista.

Fiji?

[Bob54321]

I didn't know outsourcing had extended there. Apart from the occasional disagreement between the government and the military, it seems a nice place to take a job...

Re:Fiji?

[MichaelSmith]

occasional disagreement between the government and the military

Is there a difference between them this week? I haven't been keeping up.

I don't think they get it

[value_added]

Windows 2000 promised administrators the ability to manage everything from the command-line. That turned out to be true mostly for a small list-of-old-DOS-utilities value of true. Additionally, we were offered junctions/mount-points which sortofkindof worked, but weren't fully supported. Sysinternals offered their 'junction' utility which worked a bit better, but again, not really. Now with Vista have SFU or SFU-as-subsystem that promises everything that Windows Scripting Host promised and more!

I expect that whatever hodge-podge of new features, one-off Resource Kit utilities or whatever else Microsoft decides to offer in their latest and greatest, I'll continue to rely on the folks at Cygwin to take advantage of whatever limited functionality exists in Windows, and then implement workarounds for the inconsistencies and shortcomings to make something useful and sane with it. In the meantime, I'll bet my right monad that a future Slashdot headline will read Vista's Borked NFS Client.

Re:I don't think they get it

[Anonymous Coward]

Check powershell (aka monad), it's RTM. Maybe finally you will be happy.

Re:I don't think they get it

[ZenShadow]

Last I heard, PowerShell doesn't work on Vista...

--S

Re:I don't think they get it

[pe1chl]

And not on Windows 2000 either...

Re:I don't think they get it

[Anonymous Coward]

To be fair WMIC has improved the possibilities to manage Windows from the command line by quite a bit. It still has a long, long way to go though.

Vista Symlinks not accessible via in XP and 2000

[Yahma]

People are asking questions about VISTA Symlinking on MSDN. See this thread. The Vista symlink seems to have not much more functionality than "shortcuts" did in Windows 95 or Windows 98.

The issue at hand is why was the API left so incomplete that remote accessing a share that utilizes Vista Symlinking does not work? This is a large oversight on Microsofts part, and basically makes Symlinking useless. Fortunately, Symlinking works great via Samba. Another reason to stick with Linux..

Yahma
ProxyStorm - An Apache based anonymous proxy service for security minded people.

Holy negativity Batman!

[Anonymous Coward]

Broken this, limited that. What the heck? They're IN Vista! Isn't it about time? I, for one, welcome our almost-functional---er, partially-functi... er, soon to be functional post-post-post service pack 4 Vista SymLink overlords!

Seriously though. I at least give them credit for trying.

Re:Holy negativity Batman!

[jpardey]

Yeah, they may be a huge corporation with some of the best computer scientists in the industry, and you would expect them to be able to get it, but even though they don't, IT SHOWS GOOD SPIRIT THAT THEY TRIED.

Re:Holy negativity Batman!

[Anonymous Coward]

Please. With a track record like their's, trying is about all you CAN give them credit for.

Re:Holy negativity Batman!

[AHuxley]

Link, or do not link. There is no 'try.'

Re:Holy negativity Batman!

[Sardaukar86]

Rubbish - a corporation with large financial and technical resources such as Microsoft doesn't half-implement a simple OS concept like symlinks because they are simply unable to do it properly.

It's their OS, they know all the dirty little secrets of their code and they can make it happen if they want to. Rather, I suspect it doesn't suit them to have a completed api at present.. in fact I'll even hazard a guess that (unsurprisingly) their motivations in this matter will be less to do with product quality or customer satisfaction and more about whatever FUD campaign is currently coming up to the boil in Redmond.

Re:Holy negativity Batman!

[Jugalator]

IT SHOWS GOOD SPIRIT THAT THEY TRIED

Sorry, but that use to be the kind of patting on the back you give to children who come last in a competition. :-)

Re:Holy negativity Batman!

[whoop]

As we learned several weeks ago in Clinton's rant on Fox News Sunday, "at least I tried." That's what's important in the big picture.

Re:Holy negativity Batman!

[Anonymous Coward]

IT SHOWS GOOD SPIRIT THAT THEY TRIED

Sorry, but that used to be the kind of patting on the back you give to children who come last in a competition. :-)

Yeah. That would be... the point. Of his joke. Ahem. Yeah.

Re:Holy negativity Batman!

[Neoncow]

As we learned several weeks ago in Clinton's rant on Fox News Sunday, "at least I tried." That's what's important in the big picture.

In proper context, Clinton was comparing himself to the current administration who, as he asserts, DID NOT try at all (until September 11).

Re:Holy negativity Batman!

[plopez]

Rubbish - a corporation with large financial and technical resources such as Microsoft doesn't half-implement a simple OS concept like symlinks because they are simply unable to do it properly.

Having worked for fortune 500 companies, I would say that you are giving them more credit for wisdom than is usually the case.

The environments I have see they usually:
1) Have a host of managers who have a poor grasp of what they are managing

2) Tend to go with the lowest bidder when contracting out for services, e.g. outsourcing programming and IT teams. This means monkeys usually end up doing thier software development and IT work.

3) Are run on an industrial paradigm, even in service industries. A 'let's ship tons of crap to make money' while ignoring the quality of the crap.

4) Managers make decisions based more on ego than business cases.

5) Treat people as commodities, leading once again to the hiring of monkeys to get critical tasks done.

6) Anything flash, regardless of utility takes precedence over the less sexy but often import details. E.g., nice new eye candy for Vista but a half-assed file system.

I have also worked in SME environments. I have observed that the larger the company, the dumber it gets. Larger compaies have plenty of places where dead wood can accumulate, can lose money at a rate that would sink an SME it short order and attract politcal people who know how to shift blame.

I have helped my current employer, an SME, grow to the point where it is getting too big for me. I am seeing the dysfunnctional aspects begining to develop. It is time for me to move on.

But all-in-all just because the company is big does not mean it possess an special wisdom. Esp. a monopoly which does not have to compete.

Re:Holy negativity Batman!

[Fulcrum of Evil]

It's worse than that - they exerted effort to axe programs aimed at fighting Al Queda. If they had just left well enough alone, 9/11 may not have happened.

fud

[Anonymous Coward]

FUD. Windows already supports Junction Points. RTFA. The Vista implementation of symlinks is NOT BACKWARDS compatible with older OS versions. When will Slashdot decide to publish a story that doesnt serve to garner adviews only?

22..

[psavo]

22 bloody years...

<nelson>haha!</nelson>

Not the only thing that can't be accessed...

[Karem Lore]

You website has also said goodnight...Maybe it's running off of Vista...May we suggest, erm, Linux?

Junctions

[wandazulu]

I can't RTFA because of web error, but while I haven't tried Vista's idea of symlinks, I have used junctions, which were introduced in Win2000. To me, symlinks are one of the best features of Unix and on my Mac and Linux machines, I use them quite extensively. On Windows, while the junction API was available, no Microsoft-specific tools made use of them (that I could find), and resorted to a freeware program that implemented the junction api.

Whoa, big mistake. Junctions *do* work, but, and I think this is why Microsoft didn't promote or encourage their use, none of their other tools support them. In other words, doing a search of a drive that has junctions can lead to infinite recursion depending on how the junction is created. No Windows tools understand the "Don't follow symlinks" command that Unix tools have, and I had a few programs even crash whenever I tried to save to a junctioned-folder (Visual Studio was guaranteed to crash on me).

Re:Junctions

[pe1chl]

And when you copy a directory tree that contains a junction, then delete that copy in the explorer, you will delete the original.
Very clever!
Even more so because there is no warning about this, or any way to see (in the explorer) that this is going to happen. E.g. because of a different icon or color used for a junction, like for a hidden file or a compressed file.

This stupid explorer is warning me that I want to delete a read-only file (as if anyone cared), but it smoothly deletes data it should not touch.

Fortunately, the only junction in most systems is in the SYSVOL directory. Just remember to NEVER copy that!

Re:Junctions

[dwarfsoft]

There is also the great problem that you have multiple junction points, like a School Build I know of, that gives junction points to each user. Users like to rename these whatever, and when it comes time to migrate the data across to a new system you end up copying repeated data (at least if you use xcopy, which we did at the time).

I had to write a script that read through every folder, and every subfolder on the drive, identified junction points and killed them. It still didn't work if the junction point had certain special characters in it, but killed enough of them so that I didn't end up over filling the new Server.

I'm surprised that ntbackup even allowed you to disable backing up the contents of junction/mount points. What a nightmare.

Re:Junctions

[Anonymous Coward]

Hardlinks on XP are equally retarded. You can create them using CreateHardLink() and they work generally fine. However, if you hardlink a read-only file you find that you can't delete the links to the file without first making it read-write. Unfortunately making the file read-write does so for all links, which isn't what you'd expect from Unix experience, and generally isn't what you want either.

Re:Junctions

[pe1chl]

Unfortunately making the file read-write does so for all links, which isn't what you'd expect from Unix experience

This is what you expect from Unix experience! The only thing that can be different between two hardlinked files is the name. All the attributes are shared.

Re:Junctions

[gbjbaanb]

Read very carefully. NTFS junctions are the equivalent of unix hardlinks. NOT symlinks.

I use the junction tool too (so I can support 2 versions of my app), VS2005 works perfectly well with it, but I keep everything local, and as simple as possible. I've never had a problem with junctions.

Re:Junctions

[spitzak]

You are confusing Unix hard and soft links. The behavior you specified is what Unix hard links do, which is why they are a pain and useless (except for low level stuff like atomic file rename/move). Everybody uses softlinks. Microsoft refuses to make a working softlink because it will make it too easy to make Windows compatable with Unix.

Re:Junctions

[Anonymous Coward]

Na - on Unix you can rm -f the link without having to first change the attributes. On Windows (XP at least), you have to make the file read-write before removing a link with del. This is a right old pain if you only intend to remove one link to a file and not change its attributes. The API code to find hardlinked files and fix up the attributes after the delete, while not impossible, is fiddly and would require re-implementation of del.

Obligatory rebuttal

[Tim C]

"A witty saying proves nothing." - Voltaire.

Re:Obligatory rebuttal

[Mr2001]

"Bullshit! Witty sayings are the very essence of wisdom, and I should know." -- Oscar Wilde

Re:Obligatory rebuttal

[Anonymous Coward]

"You're just making these quotes up, aren't you?" -- James T. Kirk, stardate 44094.5

Re:Obligatory rebuttal

[ettlz]

"Uncyclopedia doesn't count." — ettlz

Re:Obligatory rebuttal

[Kopretinka]

"93% of the top 640 well-known quotes are made up." -- William Gates III

Re:Obligatory rebuttal

[Anonymous Coward]

"Voltaire can suck my balls!" -- Paul Finch

Re:Obligatory rebuttal

[EnsilZah]

"People will take anything you say seriously if you put a famous name next to it." -Albert Einstein

Re:Obligatory rebuttal

[happyhangone]

"Suck my dick." -Ron Jeremy

If this is the worst ...

[Marbleless]

... that /. can find about Vista then Microsoft have won.

Windows' FS / the alternate universe story

[drgonzo59]

Window's file system has always been screwed up. Too many unnecessary attributes (remember all those times when you tried to delete the file it won't let you, saying it is in use), then drive letters are just silly and bunch of other problems.

The legacy software is both a blessing and a curse for Microsoft. They cannot break legacy code but Apple gambled and did it, and I think it succeeded. Subjectively I know more and more people who ditch their Windows machines for Macs or even Linux.

Now imagine in a parallel universe where MS decided that instead of NT or Windows XP they would just base their next OS on a compatible freely available Unix system (a *BSD for example ;). They probably would not have had to work as hard as they did writing a new NT kernel from scratch. We probably would not have had Macs and Linux-es today like we do. Microsoft could have released very cheap versions of it's OS for home and personal use ($35/pop so that the most avid pirate would rather buy it than waste time downloading it, and to make sure every child grows up immersed in a windows environment). They would have a great (free) kernel that would be secure, a powerful network stack that doesn't suck and a almost default free server that they could offer their corporate clients. Instead of supporting old legacy software they could have had some virtualization or emulation more for old DOS programs and so on. If that had happened 10 years on they would have had total monopoly in the OS market.

Today the best move MS could make is probably embrace Linux(Unix) with all they've got. They could make their own free *nix distro (yeah, I know Novel's Suse...), make it easy to install and run Windows software on that (something of a fast and reliable Wine). I bet a lot of the Linux crowd would jump ship quickly if they could play DVDs, mp3's, run Photoshop and Office on it, while still having their command prompt, the network stack and ext3/XFS etc. But Windows is still hesitating, it is a giant that has got a huge momentum and can't just stop and turn in the Unix direction, although it sure is starting to look in its direction...

don't use NTFS

[drgonzo59]

Ext3 has great symlink support ;)

Try this and a ext3 file system. I have all my Documents and the whole user directory on an ext3 and it works great. I can also access it from Linux if I want...

Re:don't use NTFS

[mrchaotica]

Is it possible to run all of Windows on EXT3?

Re:don't use NTFS

[swarsron]

Can you give us an impression of the performance? I found nothing in the docs on this page

Re:don't use NTFS

[stinerman]

AFAIK, XP refuses to boot from anything other than NTFS or FAT32(/16?). The NTLDR probably doesn't recognize anything else.

Re:don't use NTFS

[drgonzo59]

Subjectively it works great for me, I copy large media files and visibly it works no slower than a corresponding NTFS or FAT partition did. However I did not benchmark it exactly so it could be slower. Shoot the author an email and ask him, nobody knows the internals of the driver (and the possible speed penalties) better than them.

The only thing I would be worried is corruption not speed. I have never had problems, but I would not put a financial database on it either, just because it is somewhat new and "experimental"...

Re:don't use NTFS

[l3v1]

my Documents and the whole user directory on an ext3 and it works great

Good that it does, but you shouldn't advertise it as being true ext3 since it isn't. The ext2 ifs windows thing doesn't support ext3 journaling, it just treats the ext3 volumes as if they were ext2.

Point me to a true xfs windows driver and I'll be happy.

Re:don't use NTFS

[rucs_hack]

there's a big difference between 'could' and 'allowed'.

I doubt that NT or any other windows would have much trouble on ext2/3, but microsoft are not exactly likely to allow it.

Re:don't use NTFS

[EvilIdler]

Permissions would be a problem if you put all your data on an ext3 disk under Windows, too.
The on-disk and network permissions are closely related, I think.

Re:don't use NTFS

[alan.briolat]

At the risk of being modded a troll, I will say this... With the frequency at which many Windows installations need to be hard rebooted, is an XFS driver for windows a good idea? I have personally found that XFS does not handle a push of a reset button well, in my case having overwritten every open file with 0's.

Re:don't use NTFS

[b0s0z0ku]

Permissions would be a problem if you put all your data on an ext3 disk under Windows, too.

XP, at least, will run under FAT32 which basically lacks permissions and ACLs. Couldn't it, in theory, work under ext3, assuming that it ignored anything but the first +r flag?

-b.

Re:don't use NTFS

[ZERO1ZERO]

My headless file server is running xfs and on occasion when I accidentally forgot to shut it down remotely, before switching off my main PC, I have just yanked the cord. Never had any problems. (touchwood)

Re:don't use NTFS

[flyingfsck]

Nope, not quite: "The driver does not allow accessing special files at Ext2 volumes, the access will be always denied. (Special files are sockets, soft links, block devices, character devices and pipes.)"

Re:don't use NTFS

[alan.briolat]

I too have never had problems yanking the power cord - however, in my experience using the reset button is a "harder" reboot than yanking the power cord, causing an instant interrupt of everything instead of using whatever latent charge is still in the capacitors.

Re:don't use NTFS

[l3v1]

personally found that XFS does not handle a push of a reset button well

Might happen, since it just journals metadata, but it's not alone with this (reiser and jfs also, IIRC). Good thing that I use xfs for many years now and no data loss up to now, despite having been through some power outages some resets and no UPS.

BTW, crossmeta's http://www.crossmeta.com/crossmeta.html xfs mounting seems to work, more or less, I only dared to try it in readonly mode.

Re:don't use NTFS

[Hal_Porter]

NTLDR is a pretty special piece of code.

http://www.comptechdoc.org/os/windows/ntwsguide/nt wsbooting.html

214432 bytes contains read only versions of NTFS and FAT. Oddly enough, if you have a SCSI card which doesn't have Bios support, it can use the normal NT driver if you rename it NTBOOTDD.SYS, so it must contain stubs for big chunks of the kernel and IO subsystem, presumably in single threaded mode. On 99.999% of PC's it has a built in driver which switches back to V86 mode to use the Bios to read sectors. It also knows how to understand the registry, so it can load the device drivers which are needed ultra early in the system boot.

But adding support for another filesystem inside that is basically impossible. I'm suprised that it's not NTFS only to be honest, since they could tailor NTFS to make the boot process simpler, whereas FAT was a pre existing standard.

How does this fit in with...

[Anonymous Coward]

the recent news about Microsoft trying to threaten Linux users? If I'm not mistaken, symlinks were available in the Unix/Linux world long before Microsoft's sad attempts to implement them in the Windows world. Doesn't that mean that Microsoft is copying ideas from our back yard.

Only on Slashdot...

[StarkRG]

Only on Slashdot can a first post be redundant...

Re:Only on Slashdot...

[Anonymous Coward]

... and insightful

Re:Windows' FS / the alternate universe story

[jt2377]

"Linux going to take over the desktop" - 1999

"Linux going to take over the desktop" - 2002

"Linux going to take over the desktop" - 2006

the never ending ranting of slashdotter bitching and moaning.

Re:Windows' FS / the alternate universe story

[rynoski]

just base their next OS on a compatible freely available Unix system

Why this obsession with UNIX?

There are plenty of other OS's they could base thier OS off, maybe even VMS.
Unix is far from perfect, I want choice but when Linux fanboys here choice they think it must be a choice which distribution you use, because Linux is the only choice.

Back to the line about VMS, because NT was built by a bunch of ex-DEC guys, the NT Kernel isn't that bad.

I mean, they could always port GNU userland over to the NT kernel, but dont MS already do that (or something similar) in their UNIX resource thing, which you can download.

Don't worry, it isn't.

[mrchaotica]

There's plenty of other worse things about Vista; this is just an amusing side note.

Huh? Symlinks were in XP/2000 NTFS all the time

[LanceUppercut]

What are you talking about? Both Windows 2000 and Windows XP already have built-in support for symlinks (NTFS 5.0). What does all this have to do with Vista?

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[Computer Guru]

No they don't. They support "Junctions" which are the equivalent of Unix "Hardlinks." Symlinks add "Softlink" support to Windows. See http://neosmart.net/blog/archives/278 for more info.

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[ozzee]

At first blush you would think so, but the semantics of a windows NTFS symlink is more like a mount point.

Firstly, you can't use normal "commands" to create/delete NTFS symlinks. The collection of system calls to create a symlink are badly documented and incomplete.

The big differences are (from memory):

• If you attempt to delete an NTFS symlink using normal "remove" system calls, you delete the file being linked to.
• If you rename a symlink, you're renaming the file being linked to.
• There are no "relative" symlinks (i.e. ln -s x y)
• In *nix, if xxx is a symlink to a directory then xxx refers to the symlink while xxx/ refers to the location linked to
• ... probably more, I can't think of anything else right now, I'll go do some more research later.

So while there is a similar function called a "reparse point" it is more like "mount" that it is like a a true UNIX symlink.

Also, most of the Windows tools just don't deal with link (hardlinks or symlinks). If you do a hardlink from one file to another and check to see disk usage of a directory, the file is counted twice. Most unix tools (du etc) will not count hardlinked files twice and symlinks are counted as the disk space the symlink uses not the file it points to.

Symlinks are a very powerful tool and are very mature in the *nix world. Windows is just simply way behind on this one.

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[ozzee]

NTFS has supported hard links as distinct from "Junctions" and "Junctions" are not really hard links, they're just very bad implementations of a symlink which is really closer to a mount point, because thats what junctions are meant for originally anyway.

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[LanceUppercut]

You are talking about hardlinks. Hardlinks for files and directories were in Windows even earlier. They are supported since NTFS 4.0. In XP/2000 NTFS 5.0 added support for symlinks for directories. I never herd about anything new in Vista in this regard.

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[Foolhardy]

No, ozzee knows exactly what he's talking about. NT 5 suppports filesystem "reparse points" which are closer to symlinks than hardlinks because they contain a pointer to the target file in the form of its filename.

When you try to open a file (or directory; just a special type of file), the filesystem checks to see if the file (or any of its path elements) is a reparse point. If so, the filename you're searching for is replaced with the target of the reparse point (a new filename), and the kernel starts a search for this new filename from scratch. If the open succeeds, the file you have open is the one targeted by the reparse point; the fact that it was opened via a reparse point is forgotten. It's the same as if you had specified the reparse target in the first place.

Let's say that you have two disk volumes: C: and D: which Win32 has pointing to \Device\HarddiskVolume1 and \Device\HarddiskVolume2.
A program tries to open C:\rp\f.txt.
Win32 converts this to \Device\HarddiskVolume1\rp\f.txt
The IO manager creates a file object F named \Device\HarddiskVolume1\rp\f.txt and asks the the filesystem mounted at \Device\HarddiskVolume1 to find it.
The filesystem realizes that \rp is a reparse point to \Device\HarddiskVolume2\dir.
The filesystem replaces F's name with \Device\HarddiskVolume2\dir\f.txt and returns F to the IO manager.
The IO manager starts the file search over again, this time asking \Device\HarddiskVolume2 to open \Device\HarddiskVolume2\dir\f.txt.
\Device\HarddiskVolume2 finds \dir\f.txt and the open operation succeeds.
As far as anyone knows, the program had asked to open \Device\HarddiskVolume2\dir\f.txt directly, because that is the name now associated with F.
If the program wants to delete/rename the file, it's now referring to D:\dir\f.txt, not C:\rp\f.txt.
If the program had really intended to open the reparse point itself, it had to specify FILE_FLAG_OPEN_REPARSE_POINT during the open operation. The Win32 DeleteFile() and MoveFile() (for rename) functions do not specify this flag, and so affect only the target of a reparse point.

In all the versions of Windows I know of, NTFS will only let you create directory reparse points, and the target must be to a local volume (not network). AFAICT, there are no technical reasons for these restrictions. I think a lot of the concerns are about how many apps would be incompatible with full symlinks.
At some point, I'm planning to write a kernel driver to try and bypass this check. Even if I couldn't get NTFS to use non-direcory, arbitrary target symlinks, a filesystem filter could intercept the open requests and enact reparsement itself.

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[Foolhardy]

Firstly, you can't use normal "commands" to create/delete NTFS symlinks. The collection of system calls to create a symlink are badly documented and incomplete.

There are no dedicated syscalls for manipulating reparse points. To create a reparse point, you create an empty file or directory and then use the FSCTL_SET_REPARSE_POINT with the NtFsControlFile syscall (possibly via the DeviceIoControl Win32 function). To open an existing reparse point, specify FILE_FLAG_OPEN_REPARSE_POINT during a normal open operation, use FSCTL_SET_REPARSE_POINT to get its properties and FSCTL_DELETE_REPARSE_POINT to make it not a reparse point anymore. All of these operations are documented.

As for tools, the Windows resource kits have linkd.exe, mountvol.exe, and delrp.exe (mountvol is standard since XP), and Sysinternals has junction.exe. Junction comes with source code. However, there is no GUI for reparse points, and as you pointed out, lots of apps are not prepared to deal with them.

Re:Huh? Symlinks were in XP/2000 NTFS all the time

[loyukfai]

I have a question regarding mounting NTFS partition as a sub-path. Basically, I have an external drive, and I mounted it under \My Documents for faster access.

However, I'm not so sure how programs treat such kind of relationships, I've configured Google Desktop and Picasa NOT to index the sub-path but to index the drive letter directly.

Do you have similar configurations...? Any suggestion for a better setup...?

Re:Windows' FS / the alternate universe story

[MichaelSmith]

about VMS, because NT was built by a bunch of ex-DEC guys, the NT Kernel isn't that bad.

Back when I worked in VMS I got invited along to the local DEC HQ and recieved a lecture from some MSFT guy on how much NT is like VMS and that we should all run NT on alphas.

He asked for questions and I ask him why NT doesn't have proper vms-esque device names (dka0, etc). The question didn't go down very well. I supose that would break too much stuff.

I mean, they could always port GNU userland over to the NT kernel, but dont MS already do that (or something similar) in their UNIX resource thing, which you can download.

I can't imagine them working on GPL'd stuff and having to release the code.

How is this different from Unix?

[Anonymous Coward]

I haven't read TFA because it's unavailable, but as far as I can tell the implementation is exactly like it is on Unix.

On a Unix file server, symlinks have to be interpreted by the client. If your NFS client doesn't understand symlinks, it won't follow them. Of course since symlinks are new to Vista, only Vista will understand them. It is very bad for security to have the file server follow symlinks for you because it would enable you to link to files that would otherwise be inaccessible.

Likewise, it is a security risk to have a web server follow symlinks (at least by default). I suppose there could be an option to allow it (as there is in Apache), but it's such an obscure feature that the benefit of putting it in is almost certainly outweighed by the security risk.

dom

Re:How is this different from Unix?

[I'm Don Giovanni]

What you are saying is generally the consenses at osnews.com (having read their version of this thread).

Who cares?

[Anonymous Coward]

Subversion doesn't support symlinks either and everyone is desperate to bin CVS for it.

Don't underestimate Vista

[Inaffect]

I don't think its right to just attack Windows Vista because of a few problems. I mean it is not perfect and will not behave exactly the same as Linux, and this is Slashdot, but Microsoft, as a commercial entity, has done a good job providing a quality OS for computer users. Microsoft, no matter how much you may think they've "lied, cheated, and stealed" their way to the top, only has the profitability and money they have due to consumers. And they do make legitimate attempts at patching their OS and working towards improving it. I've also had about 6 Coors Lights tonight since its the weekend.

Re:Don't underestimate Vista

[Sinryc]

You know, I agree with you completely. If we can get something that is easier, faster, and all in all better, people will use it. Look at Firefox, will ya? Also, I am STILL nursing a hangover, it being the weekend and all.

Re:Don't underestimate Vista

[theurge14]

Proof: the more you drink, the more attractive Windows becomes.

Re:Don't underestimate Vista

[Vexorian]

Is this post actually an attack to vista though? It seems to talk about Vista's symlinks only, and they are limited. It doesn't make any sense to say "Symlinks are limited but the other vista features are so cool it doesn't matter" in an article that's dedicated to symlinks, does it?

Re:Don't underestimate Vista

[marcosdumay]

Hey, does nobody else see the parent as a +5 funny? Even better if you have mod points...

Re:Don't underestimate Vista

[Jesus_666]

Maybe it's not right to continually bash Vista for a few problems, but Vista doesn't have a few problems - it has heaps of them. It's late, most of the features we were promised are missing and others are badly implemented - in this case they promised us full-blown symbolic links and in reality delivered a functional equivalent to Windows 95's link files.

I'm sorry, but much of the continual bashing of Vista comes from the fact that we're continually discovering new flaws. It's hard to stop bitching when you're subjected to a continuous stream of news saying that "Vista feature X was dropped/doesn't work as advertised/is implemented in a bad fashion".

Re:Don't underestimate Vista

[Anonymous Coward]

Microsoft, as a commercial entity, has done a good job providing a quality OS for computer users.

Could you give some more information about this quality OS? My workplace is a "Microsoft only shop", and I would really like to get rid of Windows, and replace it with a quality OS.

Wait, you're not talking about Xenix, are you? MS sold that piece of junk to SCO years ago.

Re:Windows' FS / the alternate universe story

[Lord Kano]

They cannot break legacy code but Apple gambled and did it, and I think it succeeded. Subjectively I know more and more people who ditch their Windows machines for Macs or even Linux.

Apple didn't have as big of a catalog of software to abandon.

I bet a lot of the Linux crowd would jump ship quickly if they could play DVDs, mp3's, run Photoshop and Office on it, while still having their command prompt, the network stack and ext3/XFS etc.

I just played a DVD from Kubuntu's live incarnation yesterday. I have been running MP3s on Mandrake and RedHat for at least 7 years. Photoshop is still a huge hurdle. Office alternatives (especially for home users) have been plentiful for years.

LK

Re:Windows' FS / the alternate universe story

[Kaetemi]

drive letters are just silly

You don't really need to use them. In Windows you can perfectly access the volume where you installed it by simply using "\". You can also mount a volume to a folder, Control Panel > System and Maintenance > Computer Management > Disk Management > Right Click a Volume > Change Drive Letter and Paths > Add > Mount in "\blahblah" (make sure you first create the folder) > OK, and then you can just access that volume without using drive letters by doing "\blahblah" (at least, it seems to be working that way here, on vista), and you can also access these volume folder thingies from the network if it or it's parent's folder is shared. :)

If they would have kept their original ideas

[Iloinen Lohikrme]

I remember back from the beginning of 90's, around the time when Windows NT 3.11 came to markets, that vision behind NT was that it would be as modular as possible and allow swapping of lots of components beginning from the kernel to file-system. This was actually reported in lots of computer news papers, but it seems from now that it was just hype and hopeful wishes. Now it seems that the code base of NT and it's successors is so mingled that trying to swap components from it would make the system die in a split second.

Re:If they would have kept their original ideas

[RAMMS EIN]

There were very many good ideas about NT. Microsoft either actually were planning to deliver these features, but somehow didn't manage, or they deliberately misled people so that these people would wait for NT, instead of buying a competing OS. Similarly with Vista: they promised many great features, but almost all of the features I've seen people raving about have been dropped in the meantime.

Re:If they would have kept their original ideas

[badfish99]

At the time, it looked as though many commercial customers (especially the US government, as I recall) would insist on posix compliance. Also, MS wanted to appeal to customers wanting OS2. So NT was indeed designed with posix and OS2 emulation modules, as well as a Windows API layer.

Of course this was just for marketing purposes: once the customer was baited into Microsoft lock-in they would discover that this compatibility stuff was all half-backed and buggy, and would switch to using the Windows API instead.

Re:If they would have kept their original ideas

[Knuckles]

Microsoft either actually were planning to deliver these features, but somehow didn't manage, or they deliberately misled people

My bet: Engineering came up with great ideas and those were communicated early in development. Then the Marketing and Lock-in departments cut the features left and right. For example because the ability to easily swap kernel or file system does benefit the customers and makes it harder to enslave them.

Re:If they would have kept their original ideas

[TheRaven64]

Most of the nice features of NT died either because they came with a performance penalty (e.g. user-mode display drivers), or because they didn't integrate nicely with win16, and later win32 (e.g. multiple API subsystems). It still keeps some nice features; the NT kernel was designed on the assumption that system calls should be pre-emptible, all objects have an associated ACL, but modern NT-based systems bury the kernel under so much crap that it's not worth the effort.

Re:If they would have kept their original ideas

[grcumb]

Most of the nice features of NT died either because they came with a performance penalty (e.g. user-mode display drivers)

I remember when Microsoft announced that they were moving graphics back to ring 0 with NT 4.0. The rationale was that they would get much improved performance, which is a perfectly fair conclusion. But when asked about the liability created by this, they basically said, 'Well, if your graphics driver is hosed, you can't use the system anyway, so you'd just as well reboot.'

That wasn't the last straw for me, but it demonstrated with perfect clarity that this was not the OS I wanted on my servers. Here was Microsoft telling me with a straight face that they had no plans to ever provide any decent remote control of the server, that multi-user scenarios were off the table for the foreseeable future, and that system performance was going to be compromised in order to draw windows more quickly, rather than optimised over time in userland.

It was clearly a fundamental technical design decision made by the marketing department, who apparently would be happy to put tits on a bull if it increased market share in the hermaphroditic bestiality demographic.

Within two years of that, I was working exclusively with FOSS on my servers, and have never looked back.

There's a lot of mindless partisanship on Slashdot, but I think it's useful to remember from time to time that many posters here have learned to dislike Microsoft the hard way: through bitter experience of watching marketing continually triumph over technology.

Re:If they would have kept their original ideas

[Anonymous Coward]

Here's what I think happened..

          If you look on LKML (Linux Kernel Mailing List) or predecessors, during development series of kernels there will be these big raging flame wars over changing or adding some feature. Often times, changing/adding a feature, it is apparent that one overall code design could be up and running relatively quickly, while another design might be harder to get all up and running, but be more featureful, more elegant, or both... so, sometimes the quick way is chosen just to get things going, sometimes the elegant way, sometimes both in parallel to get a feature running, with change to the elegant solution later on.

          Well, with Windows design, you also have the pressure to release on schedule. So, I would guess you'd pick the quick to code way more often. So, if it takes a little more effort to keep things portable and modular you'd then tend to move away from this.

The problem is replication

[SuperKendall]

Why this obsession with UNIX?

I think that was more an example of something they could have done - basically a good choice, even though there are others.

The main problem I see with Microsoft is the incredible degree to which they duplicate something that already exists:

* Operating system, as noted they could base this on UNIX or something else and saved a lot of effort.

* Filesystem - why does the world need NTFS? There are other really good file systems around. If it offered features like ZFS I could see it but about the only FS I'd like to use less than NTFS is FAT, and that's actually a better choice for small devices because it's simpler!

* Display format - PDF ain't good enough for Microsoft, hell no, we need a brand new document/display language. Metro!

* Porgramming langauges. We can't extend Java just the way we like without community review? Screw you all, we're building a new ball from scratch and running home!

* I think we need an XML based document format. It's not like one already exists or anything, let's create one from scratch!

Think of how far the industry as a whole would be along if Microsoft actually contributed to any of those fields instead of devoting huge numbers of resources to creating anew. Microsoft single handedly has set the computer field back probably a decade or more.

Re:The problem is replication

[drgonzo59]

I agree. Re-inventing the wheel could be good if it is the right wheel, at the right time for the right vehicle, otherwise it doesn't make sense.

I watched a PBS show the other day where Bill Gates was taking questions from college students. One of the questions was "Who do you look up to for advice?" and he said "Nobody!" and that basically he is rich and smart enough that he _always_ just came with future directions for his company (MS) completely on his own. He talked how sometimes it failed (remember WebTV included in Windows 98?), and I thought to myself how that mentality of "we don't look at others, we are the smartest, biggest and we'll re-invent everything" has probably hurt MS just as much as it helped. Not many operating system concepts have come out of Microsoft as being better than what was already out there. There was always a better kernel, a better file system, a better network stack, a better security model, a better system API and so on. (By "better" I mean better implemented practically, sure NT has a great security model but practically it is not the best, and "theoretically" Windows XP would support POSIX...but in reality it doesn't)

Re:The problem is replication

[DannyO152]

Let's add PowerShell to the above list, no doubt a cool implementation (and maybe the catalyst for development of F# [Microsoft's under-development ML]), but why not also put bash on a tools disk and let savvy customers use their experience today (or five years ago when Microsoft noticed that command-line utilities were a feature, not a bug, and power-users were buying OS X and Linux systems when allowed to choose.)

Re:The problem is replication

[Anonymous Coward]

* Operating system, as noted they could base this on UNIX or something else and saved a lot of effort.

It's true that Microsoft had a UNIX flavor in the works called Xenix, but it was not completely owned by them (they had only licensed it) and simply because it was UNIX didn't make it good. The NT kernel is actually a magnificent piece of work, and claiming otherwise is just an insult to Dave Cutler. The quality of what's on top of it is debatable, but the groundwork is solid, IMHO.

* Filesystem - why does the world need NTFS? There are other really good file systems around. If it offered features like ZFS I could see it but about the only FS I'd like to use less than NTFS is FAT, and that's actually a better choice for small devices because it's simpler!

Are you fucking thick? NTFS came to be in 1993 with the first version of Windows NT. FAT could not possibly support the required features at that time nor was it robust enough. There were no other alternatives available at the time (bar HPFS, which NTFS is supposedly based on). What would you have had Microsoft do?

And on what basis can you claim that NTFS is a bad file system? Sure, it's not open, but how often does it break through something else than hardware failure? Not very often, compared to the ridiculous amount of systems using NTFS on this planet. It's quite fast and feature-rich. It Just Works.(TM) And that's a really important point to most people.

Re:The problem is replication

[SuperKendall]

The NT kernel is actually a magnificent piece of work, and claiming otherwise is just an insult to Dave Cutler. The quality of what's on top of it is debatable, but the groundwork is solid, IMHO.

If you'd read what I said, not once did I mention anything about the QUALITY of what Microsoft produced. NT is indeed an impressive product.

But again, it is done from scratch when they could have simply improved upon existing systems, with less effort and greater return to the community of computer users at large.

Similarily C# is a good programming langauge but the breathtaking waste of man hours over simply working within the JCP to improve Java is staggering, and inexcusable. If there is a programmers hell every single person at Microsoft involved in C# is going directly there.

Are you fucking thick? NTFS came to be in 1993 with the first version of Windows NT. FAT could not possibly support the required features at that time nor was it robust enough. There were no other alternatives available at the time (bar HPFS, which NTFS is supposedly based on). What would you have had Microsoft do?

You gave the answer to your own question (along with the pointless swearing, a patented trademark of the MS zeolot along with posting AC so the OP won't see, and correct, your reply) - improve upon HPFS in conjunction with other companies, and have a true cross OS filesystem standard. If that has been done I would wager there would be a good change that things like CF cards would be based on HPFS today instead of FAT and FAT32.

It's not like Microsoft is incapable of collaberaton, it's just that they seem to have an irrational dislike of doing so, and this has harmed the computer industry through duplication of effort and also termination of effort, when MS was already involved in a space and no others could aquire funding for advancements.

Re:The problem is replication

[mattgreen]

I hope you are joking with regard to C#. Have you used both languages? Java is not the end-all, be-all of object-oriented languages! Given the choice to use either Java or C#, I'd take C# every single time. The .NET Framework is superior to Java's in terms of organization and naming conventions. (It is not as deep as Java's third party support, however.) I consider Visual Studio superior to any of the sluggish-feeling-type-type-oh-wait-now-I've-gotta- garbage-collect-so-I'll-stall-for-awhile Java-based ones that everyone raves about for Java. C# is more expressive than Java, as well. I use Java daily, but at times it feels like the language has been greatly watered down so it would be the next COBOL.

You fail to realize that MS doesn't care one way about being interoperable. Why would they want a cross-platform filesystem? Their game is being proprietary. Anyway, complaining about the whole choice thing is rather silly. Isn't that what open source is all about? Frankly I'd rather have the competition to spur innovation.

Re:The problem is replication

[nuzak]

> Similarily C# is a good programming langauge but the breathtaking waste of man hours over simply working within the JCP to improve Java is staggering, and inexcusable

Sun has absolute veto power over the JCP, which is nothing more than a glorified suggestion box. If you honestly believe that Sun wouldn't have used that power at opportunity to bludgeon Microsoft, you're deluded. But I don't suspect you even considered that possible counterargument in your zeal to go after Microsoft?

I could go on about actual useful features in C# that don't have a snowball's chance to make it into Java, JCP or no, but I don't think it needs any cheerleading from me. Why is Java so "special" that it can't brook any real competition?

Re:The problem is replication

[Decaff]

Sun has absolute veto power over the JCP, which is nothing more than a glorified suggestion box.

This is total nonsense. Sun has veto (and why not - the invented the language?) but rarely use it. There are major debates in the JCP, and real innovation. It is not just a 'suggestion box' - respected industry experts help write specifications.

If you honestly believe that Sun wouldn't have used that power at opportunity to bludgeon Microsoft, you're deluded.

Defending Java against Microsoft's attempt to wreck it is not 'bludgeoning' Microsoft. In other areas Sun has worked in strong collaboration with Microsoft, as in the Java 6 + Vista integration.

But I don't suspect you even considered that possible counterargument in your zeal to go after Microsoft?

There is no counterargument here - you simply stated your belief.

I could go on about actual useful features in C# that don't have a snowball's chance to make it into Java, JCP or no, but I don't think it needs any cheerleading from me. Why is Java so "special" that it can't brook any real competition?

Anyone is open to compete with Java. .NET could well have been that competition, but Microsoft decided to keep it largely confined to Windows. Instead, a more open, free, Microsoft-supported cross-platform .NET could easily have competed. Microsoft chose not to do that.

Re:The problem is replication

[Anonymous Coward]

along with the pointless swearing

Pardon me. That was a carnal reaction to the OP's ridiculous suggestion that ZFS should have been used. (Be it from the original version of NT or Vista; in the former case ZFS didn't exist back then, and in the latter case it would have been fairly silly to venture to new lands because NTFS already exists and works and has been debugged to a large extent.)

But again, it is done from scratch when they could have simply improved upon existing systems, with less effort and greater return to the community of computer users at large.

IIRC what sealed the deal was Windows 3.0's runaway success. Up until then, they *were* improving upon existing systems - although it wasn't UNIX but OS/2. That gave way to the project to create their own 32-bit OS with support for old Windows programs (Win16), and NT was born.

Now, although IBM is the posterchild for open source and collaboration-supporting big companies now, that most certainly wasn't the case in late 80's and early 90's. IBM was the big bad company, and half of the computing world cheered Microsoft on to beat the big bad IBM. We don't and can't know the details behind the OS/2 collaboration between IBM and Microsoft, but I'm willing to bet that IBM had a much larger say in how things were to be in that project. Given that Microsoft realized they had a workforce capable of doing the thing by themselves (or were able to hire the necessary people), what company wouldn't seize the opportunity to get independent?

improve upon HPFS in conjunction with other companies, and have a true cross OS filesystem standard.

That's assuming the other companies (namely IBM) wanted Microsoft to have their hands in the project, given the already crackling OS/2 alliance.

Re:The problem is replication

[SuperKendall]

I hope you are joking with regard to C#. Have you used both languages? Java is not the end-all, be-all of object-oriented languages! Given the choice to use either Java or C#, I'd take C# every single time.

Being a copy of course they were able to imrpove on a few things. Of course they lack the depth of the Java libraries...

The .NET Framework is superior to Java's in terms of organization and naming conventions. (It is not as deep as Java's third party support, however.)

I sure hope you aren't referring to the bass-ackwords way they did the capitalization. That was an atrocity and really did away with a lot of historical conventions, for no good reason and a number of bad results.

I consider Visual Studio superior to any of the sluggish-feeling-type-type-oh-wait-now-I've-gotta- garbage-collect-so-I'll-stall-for-awhile Java-based ones that everyone raves about for Java.

Eclipse is really more powerful, and more extendable - it's the same way that Firefox is really nicer to use than IE, because there are a huge variety of plugins.

It takes more resources to run no doubt, but I find it preferrable.

C# is more expressive than Java, as well. I use Java daily, but at times it feels like the language has been greatly watered down so it would be the next COBOL.

Now that I'd quibble with, at least from Java 1.5 onward. What do you find "more expressive?"

You fail to realize that MS doesn't care one way about being interoperable. Why would they want a cross-platform filesystem? Their game is being proprietary.

Of course I realize that! That is my whole point, that approach is wasteful and duplictitive and setting the entire industry back. Look to Apple for a model, they are releasing webkit stuff back to the community along wtih GCC patches and freebsd stuff as well.

That may be thier "game" but setting people on fire can be a "game" as well, I don't care. I still consider it evil.

Anyway, complaining about the whole choice thing is rather silly. Isn't that what open source is all about?

YOU are failing to realize that open source is about the choice to do what you want with software, to extend it in ways the makers never considered and to have others benefit as well. At least that is what the GPL is about, and why there is a divide between open source advocates.

Re:The problem is replication

[larry bagina]

Similarily C# is a good programming langauge but the breathtaking waste of man hours over simply working within the JCP to improve Java is staggering, and inexcusable. If there is a programmers hell every single person at Microsoft involved in C# is going directly there.

They'll have plenty of company, with java (which was a waste of man hours over using OpenStep), OpenStep (which was a waste of an hours over using C and Smalltalk), C (which was a waste of man hours over using BCPL or B)... And don't forget linux and gnome!

Re:The problem is replication

[Anonymous Coward]

>Operating system, as noted they could base this on UNIX or something else and saved a lot of effort.

Ooh, yeah, maybe they could get Dave Cutler to be the architect and base it off VMS, which was mature before any of the Unixes even stopped sucking. Oh, wait....

Even Pike has acknowledged "Not only is UNIX dead, it's starting to smell really bad."

Re:The problem is replication

[SuperKendall]

I could go on about actual useful features in C# that don't have a snowball's chance to make it into Java,

Do please go on.

The only thing that really wouldn't make it in is the C# generics. They are cleaner than the Java version to be sure, but remember that Java had a LOT of legacy code they were trying to support as well, so they had a great reason to take the approach they did - which allows you to use generics with older Java libraries.

I have nothing against Microsoft when they do good, Office X on the Mac for example is pretty decent and I think they have brought something new to the UI table with the new ribbon interface. But C#, there is no excuse for doing the work they did to get it where it is - or indeed continuing that effort, when they themselves use it half-heartedly in the operating system.

Re:The problem is replication

[SuperKendall]

Pardon me. That was a carnal reaction to the OP's ridiculous suggestion that ZFS should have been used. (Be it from the original version of NT or Vista; in the former case ZFS didn't exist back then, and in the latter case it would have been fairly silly to venture to new lands because NTFS already exists and works and has been debugged to a large extent.)

That was more a complaint with the VIsta of today that projects of yesteryear. Yes I know ZFS has not been out that long, but how long has Sun been working on it...

That's assuming the other companies (namely IBM) wanted Microsoft to have their hands in the project, given the already crackling OS/2 alliance.

At that point IBm did have a lot of power, but at some point later Microsoft should have ceased devoting efforts to being independant at all costs, even as they grew larger than anyone... they got greedy and we have all suffered.

Re:The problem is replication

[nuzak]

It's true, C# didn't have a lot of legacy to break, while Java took an approach with generics that ensured backward compatibility at the expense of robustness. Hmm, who's been strung up before in the court of techie opinion for that? Of course C#'s library versioning means it can load the correct version of a class out of libraries in the assembly cache or path that have multiple versions of the class (such as generic-aware).

C# also has unsigned types, value types, property getter/setters, operator overloading, multicast delegates (which I'll admit abuse operator overloading). Rather incremental additions to be sure, and they don't please all the purists, but they're there.

And yes, like anything from Microsoft, it's pushed with one hand and held back by the other. I'm told that the Exchange folks have gotten the religion though, and that most of its management is done with Powershell. Wish they'd do the same for MS Office.

Re:Windows' FS / the alternate universe story

[drgonzo59]

There are ways around but most applications out there are still installing themselves in C:\Program Files and windows goes into C:\Windows and so on.

Re:Windows' FS / the alternate universe story

[drgonzo59]

There is not obsession with UNIX there is obsession with "common sense", security, stability, reliability. UNIXes at the moment fit that list most of the time. Tomorrow it could be Plan9 or QNX or some other exotic thing, I don't care. I am not a more Linux fanboy than I am a Windows fanboy or a Honda or Wal-Mart fanboy I just use what makes sense at the time (yap, sorry no brand loyalty at all here).

I mean, they could always port GNU userland over to the NT kernel, but dont MS already do that (or something similar) in their UNIX resource thing, which you can download.

You are referring to POSIX I presume. Well, have you seen any native Unix code running on Windows lately? I didn't! Windows POSIX compliance is a joke, it was more of a marketing ploy to tell their client ("we even run Unix!") but in reality it is very broken. That is why you have Cygwin...

Re:Windows' FS / the alternate universe story

[shmlco]

And what happens in that alternate universe when they do that, and suddenly are unable to differentiate themselves from dozens of other nearly identical systems? Guess they stop being an OS company, huh?

Re:Windows' FS / the alternate universe story

[drgonzo59]

That's the story of Microsoft. Everything half-ass works. POSIX is working but is also not working. Symlinks work but don't work. Security works but not quite. I think if they had done what Apple did and just used a well tested free Unix-like kernel or a complete VMS implementation with a good network stack and security they could have been further ahead today...

From Microsoft.

[Captian Obias]

"Microsoft has indicated that Windows will support true symbolic linking as of the release of Windows Vista/Windows Longhorn Server."

Re:Windows' FS / the alternate universe story

[Richard_at_work]

I can't imagine them working on GPL'd stuff and having to release the code.

As the GP suggests, the Services for Unix package contains a lot of GPLed stuff, the code for which is also available on the Microsoft FTP servers. Microsoft have no issues with actually following license terms if they are required to do so.

If you can't be better, be faithful

[dpbsmith]

Compare Mac OS X. It has two different kinds of symlinks. It has the traditional, pure-quill, UNIX symlinks which work exactly as UNIX users expect.

It also has Mac OS "aliases," introduced IIRC in System 7, which most Mac devotees think are superior to UNIX symlinks.

Now, before I get too far into praising "aliases," let me acknowledge that the presence of both mechanisms in Mac OS X is a big, hairy, ugly, mess, and one of innumerable places where the Mac world currently suffers from having anywhere up to half a dozen or so APIs for the same basic functionality. Mac OS X now resembles, well, my house, with fifteen-year-old half-abandoned dusty possessions still lurking in the attic. Not that Windows is any better, of course.

But I digress. You may like Mac OS aliases or you may dislike them, but you can see they they are a complete, well-thought-out, finished, working mechanism that it is at least possible to admire as something more than a half-baked knockoff of symlinks.

I happen to like them, a lot, because they just work. You don't need to do anything special at a programming level to dereference them, and it doesn't matter what programming language you're using or whether you're accessing them across the network, or whatever. However you do it, when you open the alias, you open the file it points to. And they are not fragile: you can move them or rename them or whatever and they still point to the right place. (The tough part is not dereferencing them... and Apple's deliberate failure to document or provide an API for creating them programmatically).

What I find hard to forgive Microsoft is that when Microsoft implements their knockoff of a well-known OS feature, it is rare that they come up with anything fresh and original. So many of their derivatives seem to be hasty knockoffs implemented by people who didn't "get" the original. And they put these half-baked implementations into shipping products, making it very difficult for Microsoft ever to finish them or fix them.

You can see this in a dozen places, like the Windows NT command language, which is a half-baked extension of the miserable quarter-baked DOS command language. Jeez, guys, you had DCL and the various UNIX shells as models, couldn't you do better than that?

And five years later, there tends to be conflicting documentation: the documentation written when badly-designed feature X was introduced, telling all good little Microsoft developers that they simply must, must, must use feature X in everything, and the documentation written a few years later warning everyone against the bad practice of using crufty old deprecated feature X...

I just wish I could shake Microsoft by the scruff of the neck and say, "Listen, if you can't improve it, then at least make a faithful copy of it."

Don't just pee in it to give it that personal flavor.

Re:If you can't be better, be faithful

[Anonymous Coward]

I don't know anything about Windows offering a serious alternative to hard or soft links. The only thing I can think of is Windows "Shortcuts". As an end user, they work just fine; you tell Explorer to "Create Shortcut Here", and from then on clicking that icon references the file pointed to.

But as a Windows programmer (who has much experience writing C/C++ code in Linux and FreeBSD), when my program is passed a "shortcut", I must then call a Windows API function to figure out what file the shortcut points to. Nothing like hard- or soft-links in any Unix-like system, where the filesystem dereferences the link automatically for you. Hell, many common Unix/Linux binaries count on this feature, and (as usual) Windows is the only OS lacking a proper symlink implementation in its default filesystem, or any filesystem Windows supports for that matter...

As mentioned above I write Windows software, only because I could never make a living selling Unix/Linux software... but here at home, I have more Linux and FreeBSD systems than I do Windows systems...

Re:If you can't be better, be faithful

[Jesus_666]

Maybe Apple should just implement symlinks as aliases (at least on HFS+ volumes). That way you can ln -s a file without the link looking ugly in the Finder and you get some more simplicity. I don't know enough about the differences between both to actually tell whether it works like that, though.

Re:If you can't be better, be faithful

[spitzak]

Yes, I agree. They should just have made the symlink api manipulate aliases. I feel it would have worked. The only real difficulty is that readlink() would probably be somewhat more complex and the result is not constant.

Currently they are a bit screwed up. They have three types of links (hard, symbolic, and aliases). And aliases, though they might be nice, have an API even worse than what Microsoft is coughing up. At least on Windows I can peek into a "desktop link" file and read the symbolic link. There does not appear to be any api for aliases!

Clarification about Mac alias robustness

[dpbsmith]

When I said "And they are not fragile: you can move them or rename them or whatever and they still point to the right place," what I meant was that you can move or rename the targets they point to without breaking the link.

Re:Clarification about Mac alias robustness

[TheNetAvenger]

what I meant was that you can move or rename the targets they point to without breaking the link.


Not to be an ass, but even freaking Shortcuts in Windows (from Win2k and newer) don't break either, even if the shortcut points to a volume half way around the globe.

Look up NTFS's DLT features...

As for the whole Vista not supporting Symlinks, people really need to do a bit of homework here. Vista supports, three main types of reference mechanisms from Symlinks to the old Shortcuts. What people are talking about is Vista's implementation of Symlinks can fail if the NTFS volume is accessed from an older version of Windows.

If anyone here actually even looked at a freaking Vista install, they would see MS has even used Symlinks in the default installation even - \Documents and Settings is a link to \Users - etc...

I get tired of this, if Windows doesn't do this trick or that trick it isn't worthy... Well lets see OSX or Linux handle multiple OS subsystems on a client/server kernel - then I'll be impressed and consider them 'worthy'. (Geesh)

Re:Clarification about Mac alias robustness

[EvilIdler]

Just like hardlinks?

Re:Clarification about Mac alias robustness

[Knuckles]

Not to be an ass, but even freaking Shortcuts in Windows (from Win2k and newer) don't break either, even if the shortcut points to a volume half way around the globe.

The parent was talking about the fact that if you have a Shortcut pointing to a file and then move/rename the file, the Shortcut points to nowhere. This is not the case with Aliases, they are updated automatically.

I'd like to add that someone at MS had the hilarious idea that if a user copies a Shortcut to a removable medium he really wants to take the 1 KB textfile with him. In contrast, Apple understands that if he copies an Alias he actually wants the file that it points to.

Re:Clarification about Mac alias robustness

[lgw]

Not to be an ass, but even freaking Shortcuts in Windows (from Win2k and newer) don't break either, even if the shortcut points to a volume half way around the globe.

Look up NTFS's DLT features...

They don't on any of my Windows boxes, presumably because DLT is one of the two-dozen useless resource-hogging services I disable. Just like you can't depend on any COM or RPC features beyond the most basic when programming for Windows, as your client will have diabled whatever service you needed. Of course, this is to be expected as the more Windows "features" one disables, the more pleasant the result, so power users diable whatever they can.

Re:Clarification about Mac alias robustness

[lgw]

The parent was talking about the fact that this is simply not true on a modern default install of Windows, even if you move the file linked to to a different machine. Just like Aliases, they are updated automatically.

Unless, like me, one hates computers that think they know what they're doing, and disables the Dynamic Link Tracking service. I want my computer to do what I tell it, not to do something arbitrary that some designer imagined that I might have wanted it to do. Windows sucks, but at least it can be coerced into following simple instructions most of the time.

Re:Clarification about Mac alias robustness

[Knuckles]

The parent was talking about the fact that this is simply not true on a modern default install of Windows, even if you move the file linked to to a different machine. Just like Aliases, they are updated automatically.

I realize that second-guessing the parent is stupid, but I don't think this is what he said. Here he replies to and clarifies himself and, speaking about Aliases as opposed to Shortcuts, says

When I said "And they are not fragile: you can move them or rename them or whatever and they still point to the right place," what I meant was that you can move or rename the targets they point to without breaking the link.

So, apparently he says that Shortcuts are fragile and moving or renaming the targets breaks the link.

As for your claim that this is not true anymore: thanks for the info. It seems then that our corporate IT changes this setting; I haven't used Windows outside of work in years.

Re:Clarification about Mac alias robustness

[TheRaven64]

Not quite. OS X actually has three kind of link:

1. Hard links point to an inode on the current volume. File names are, effectively, all hard links, although typically you only refer to them as links when the file's reference count if greater than 1. All hard links are first-class citizens; you can't say one is real, and one is a pointer, they both have the same precedence.
2. Symbolic links point to a path. If the path to a file changes, they will become invalid. If a new file appears at the path, they will point there. Symlinks can point to either an absolute or relative path. They are most useful if you want to point to a file that might not exist; for example a log file created by a program.
3. Aliases point to a file. They have two major differences with hard links. They are not all first-class citizens, there is a clear distinction between the 'real' file, and the alias. They also work across volumes. Each HFS file has a unique index, which is indexed. Aliases point to this. You can move a file between volumes, and the alias will keep working. As such, they are very useful for pointing to files on removable media; the alias will continue to work whenever the disk is mounted, even if it is on a different mount point.

Personally, I find aliases to be the most useful, conceptually. They do have the disadvantage, however, that (like shortcuts) they do not work transparently at the POSIX layer. Unlike shortcuts, however, they do work transparently if you use the Cocoa/Carbon APIs. Hopefully Leopard will move them down to the POSIX layer, as it did with resource forks and other meta-data.

Re:Clarification about Mac alias robustness

[lgw]

"Parent post" is a relative term. ;) I was commenting on the fact that you seemed to miss the bit where TheNetAvenger said

Not to be an ass, but even freaking Shortcuts in Windows (from Win2k and newer) don't break either, even if the shortcut points to a volume half way around the globe.

Look up NTFS's DLT features...

Of course, quite a few people, myself included, turn off DLT along with dozens of other services worth of Windows bloat, but it's been there on the default install for many years now.

Re:Clarification about Mac alias robustness

[Knuckles]

"Parent post" is a relative term. ;)

Right :)

I still think that the OS should do the Right Thing automatically, and if this needs a bloated service then one more thing is wrong in Windows. I understand that one might sometimes want the Shortcut not follow the moved file automatically, but having that option should be doable without breaking a good default behavior. And for the other case, I would still think that it is a safe bet that copying a Shortcut to removable media does not mean the user wants to take the Shortcut text file with him :)

Re:Clarification about Mac alias robustness

[TheNetAvenger]

Ok, I know that that when the term 'service' gets thrown around, especially in the world of Windows it sounds like a lot more than probably what people realize.

So it is true it is a service as it is implemented on Windows, but at 12K, I really don't think the word bloated would apply.

As for the user not wanting to take the shortcut/alias with them, this is a usability concept that is very subjective. I have a USB device that I have tons of shortcuts to data and folders located on servers both in house and across the world. (Both native FS shortcuts and NS shortcuts.) In my world, the Apple logic is backwards, but I am a bit more 'network' connected than most people, so I don't have to worry about the links being orphaned pretty much no matter where I am, even my car.

Re:Clarification about Mac alias robustness

[TheNetAvenger]

DLT is one of the two-dozen useless resource-hogging services I disable

You are saving yourself 1 process and 12K so that your shortcuts don't stay updated. Brilliant...

Just because something is implemented as a service, does not mean it is resource intensive. If people understood the NT kernel architecture a bit more, they would also accept why things like the DLT features are ran as a service and not jammed into the kernel for FS tracking as some OSes would do to get the same functionality.

Just like you can't depend on any COM or RPC features beyond the most basic when programming for Windows, as your client will have diabled whatever service you needed

Again, if people are dumb enough to turn off key features of the OS, they probably can't be helped, even if they are happy they are freeing another 32K of RAM. You write your software and if the service is off, you tell the user they are a moron and to turn it back on. It is pretty simple actually.

Re:Clarification about Mac alias robustness

[frankie]

Well maybe if Microsoft hadn't implemented so many "features" (that rarely get used) as services that turned out to be remotely exploitable holes, people wouldn't have such a bad impression of them. The most reliable security default for Win NT/2K/XP is "if you don't seem to be using a service, and nothing seems to break when you turn it off, disable it".

Re:Clarification about Mac alias robustness

[Knuckles]

So it is true it is a service as it is implemented on Windows, but at 12K, I really don't think the word bloated would apply.

I wouldn't know, the guy I was talking to that had claimed that people are disabling this and other services due to bloat. And as stated earlier, our corporate IT seems to do the same.

As for the user not wanting to take the shortcut/alias with them, this is a usability concept that is very subjective.

Sure, it is a question of target audience. If I were to design this for a mass-market OS however, I would make the Apple way the default without a doubt (and am pretty sure any usability test outside a comparatively tiny group of experts would support this)

[Off-topic] Your sig

[empaler]

Theory is often inaccurate(TM)

Would you care to elaborate on the usage of the word often in this sentence? Alternatively, which use of the word theory is it you refer to?

Re:[Off-topic] Your sig

[A nonymous Coward]

Theory is his dog, and it often pisses on his shoe and humps the fireplug down the street. Theory is old.

Symlinks to local dirs on shares?

[ACorvus]

So from the article, I gather it won't be possible to place a symlink on a shared directory that points to a directory on the machine that mounts the share?

We use this for a call recording system already. We mount a share via NFS on a NAS box. On that NAS box there is a symlink from a subdirectory of the NFS share to a path that doesn't exist on the server, but does on the client (the server stores the call audio, the client stores the call metadata). When the client mounts the share, the link works correctly, it points to the directory on the local machine.

This is damn useful since we can't change the behaviour of the call recorder (it expects the calls and metadata to be in the same directory). It seems this implementation won't allow this kind of transparency...

Microsoft needs a new NT

[Jesus_666]

When NT was created Microsoft knew it'd break compatibility but be superior to the existing stuff. Maybe Microsoft should treat the current NT line as a legacy OS (effectively making Visty the new Windows 98) and create a new OS completely from scratch.

I mean, how often did they have to scrap major parts (or even everything) of Vista and start over? At least twice, I think. Windows is getting unmanageable and I somewhat doubt that the NT kernel is in a better state. If Microsoft first developed a structured approach to OS building (unlike the 1000-developers-and-minimal-communication hodgepodge they have today) and then created a completely new kernel and userland parallel to Vista they might impress their users like they did when they introduced the NT kernel to home users with Windows 2000. They might even surprise people and write an OS that actually gets some respect from IT professionals.

I think Blackcomb could very well become the next Me - with or without a superior alternative being developed.

Re:Microsoft needs a new NT

[Knuckles]

create a new OS completely from scratch.

They do

Re:Microsoft needs a new NT

[Jesus_666]

Yes, but not many people expect Singularity to reach any market ever.

Re:Microsoft needs a new NT

[Knuckles]

I certainly hope it (or any other new MS OS) won't.

Re:Microsoft needs a new NT

[TheRaven64]

So, they built a cheap knock-off of CP/M, then a cheap knock-off of VMS, and now a cheap knock-off of JNode? It sounds like they're really scraping the bottom of the barrel for obscure operating systems to copy...

Re:Windows' FS / the alternate universe story

[drgonzo59]

Is OS X like the dozens of other *nix systems? Well, "Yes" and "No". It has all the core features but has enough of its own. It combines the stability, networking and security of Unix + the Apple GUI. Last time I checked, it is doing pretty well even among the other dozen or so "similar" operating systems. The point was that Microsoft could have been the Microsoft (support lots of software and hardware) and the Apple (have a stable, no-nonsense core operating system) -- it could have had the best of both worlds, in my hypothetical parallel universe.

see his home page/blog

[way2trivial]

top left corner

"I am a software developer in Seattle, building a new AI software company. I used to work at Microsoft on Excel"

Re:see his home page/blog

[Barryke]

@way2trivial: I cant find what you quote.

Re:see his home page/blog

[way2trivial]

go to link in wesnerm post
http://wesnerm.blogs.com/net_undocumented/2006/10/ symbolic_links__1.html [blogs.com]

4 lines directly under +ABOUT

Re:see his home page/blog

[Anonymous Coward]

It's in the upper left corner (the about section) of his blog at http://wesnerm.blogs.com/.

And I in no way mean for this to sound racist in the slightest, but why is it that the only people I've ever known who get involved in 'entrepreneurship' are black? (I'm referring to at least half a dozen separate instances). Everyone else I know either 'freelances' or 'goes into business for themselves'.

Re:see his home page/blog

[Anonymous Coward]

A lot of africans speak french. "entrepreneurship"

Re:see his home page/blog

[Knuckles]

Because you don't know any business consultants.

who cares?

[Anonymous Coward]

I personally don't know anyone who would want to use Microsoft's Vista. Only the few less cerebral folks among us would still waste their time obsessing with a failed attempt by Microsoft to create another Windows. Folks, wake up and use a real OS!

Why symlinks?

[twan]

User your drive letters!

"subst" working since MS-DOS 3.0

Re:Why symlinks?

[TheRaven64]

I remember using a combination of subst and join to provide something equivalent to symlinks for directories in DOS. Subst mapped a directory to a drive letter, and join mapped a drive letter to a directory. If you did:

> SUBST c:\foo\bar G:
> JOIN G: C:\bar

Then c:\bar would be a link to c:\foo\bar. Of course, it only worked with directories, not with files.

Re:Why symlinks?

[bad_sheep]

Yes... with the great limitation of 24 (or less...) links on your system (until you reach the Z: letter) !

24 links will be enough of everybody right ?

Re:Why symlinks?

[slaida1]

24 links will be enough of everybody right ?

I was laughing just the other day at the thought of MS redoing that horrible kludge in some new shiny Windows version, hyping it to heavens: "..and now we have new improved network drive mapping feature, Double Drive Letters. After letters from A to Z, there comes AA, AB, AC, all the way to ZZ."

What's Fiji

[proxy318]

hopefully this is something that can be patched via a hotfix and that we don't have to wait for Fiji to get something as simple as UNC support built in."

I must have missed something. What the hell's Fiji? (besides a small island in the south pacific)

No Compelling Reason to Buy Vista

[HermMunster]

Sheesh, stop talking about some of this banal Vista crap. Just let people know that Vista has no compelling reason to upgrade. This is a no brainer. I'm not being harsh nor trying to shut anyone down but these endless posts about little defects certain tend to color the fact that there's no reason to buy Vista, unless you want to have the latest stuff from Microsoft. If people want am OS that is solid and inexpensive that runs on a multitude of platforms then they should run Linux. Not that I am a huge Linux fan but it has to be definitely cheaper and requires alot less upgrading than buying Vista and installing it and then getting very little for the exchange of money.

So, please stop tossing little things that tend to color over the bigger issue: There's no compelling reason to buy Vista.

Re:No Compelling Reason to Buy Vista

[Malc]

I think you've got it the wrong way round. Windows requires less upgrading than Linux (meaning distro). The last major desktop release from MSFT was Windows XP, which was 2002 (actually, New Year's eve 2001). That's a longer release cycle than Debian, which is considered incredibly slow in the Linux world. Using Linux means constantly upgrading and learning new quirks and configurations. No thanks. That gets in the way of using computer for useful work. Of course, I'll carry on using XP as long as I need because MSFT will continue to support it for years. I'll get Vista when I need new hardware. MSFT is still supporting Windows 2000, which is even older still, and will continue issuing security patches for a few more years. It seems support of legacy versions on Linux is limited at best, and quite often unofficial and due to the dedication of users.

Re:No Compelling Reason to Buy Vista

[dbIII]

Using Linux means constantly upgrading and learning new quirks and configurations. No thanks.

You don't have to do it unless you are running something directly exposed to the net which you would have to patch frequently no matter what the OS is.

Re:No Compelling Reason to Buy Vista

[Malc]

That's a strange statement around here! This is /.: we're supposed to be rabid about security! ;)

Actually, I worry about anything on my LAN. Take my Tivo (series 1) for instance: I'm in Canada and cannot get any Tivo updates. I have it on the LAN because I feed it programme listings from Zap2it.com. If another system on my LAN is compromised, my Tivo is at risk. Thankfully it's not a complete Linux distro, and it's running on PPC hardware (IBM 403GCX), but it is running some services so that I can manage it and copy stuff off it. All of my full Windows and Linux boxes I run some sort of firewall on in case of a worm breaking out on the network. The only system I've had compromised (via an ssh dictionary attack, a weak password and an out-of-date kernel) was my Linux server. A packet sniffer was installed, and hidden processes opened connections to remote IRC ports. It could so easily have gone further. That's when I realized Debian doesn't automatically upgrade kernels if you don't install a specific one.

Re:Windows' FS / the alternate universe story

[Anonymous Coward]

But the issue is that Linux (or any other *nix) is just not generally usable by the average person. People want to sit down at their PC and be able to watch their movies, listen to their music and play their games without trawling websites for 2 hours, reading manuals. Even if you have to spend some time searching for the solution in Windows, chances are it will be easy to get going when you find it. Unfortunately, the same cannot be said for Linux.

Re:Windows' FS / the alternate universe story

[Knuckles]

And guess what happened: each year a larger percentage of desktops were taken over.

ext2 ifs fails to deliver

[Anonymous Coward]

From the FAQ (http://www.fs-driver.org/faq.html)

What features are *not* supported?

The driver does not allow accessing special files at Ext2 volumes, the access will be always denied. (Special files are sockets, soft links, block devices, character devices and pipes.)

Re:Windows' FS / the alternate universe story

[Anonymous Coward]

It does seem to be that every article about Vista could be prefaced with "Windows Vista will suck because..." and every second comment summarised by "I expect to see mass migration to Linux because of ". The big news is: the average user doesn't care about symlinks. The average user doesn't even know what they are, or what they do. The average user also represents the majority of the market. Until people can sit down, chuck a Linux distro on their PC and have it just do the things they want (games, movies, music) with minimum hassle and without being required to spend hours reading manuals, there will be no change in the situation.

Re:Windows' FS / the alternate universe story

[shmlco]

Apple has the luxury of marrying its OS with great hardware and design.

As pointed out, MS doesn't, and in fact MS has to support every POS software/hardware combination on the planet. Without that level of polish (which MS hasn't shown it can do anyway) and integration, MS's version would just be YAN (Yet Another 'Nix).

Vista's full symlink support is OFF for security

[Commykilla]

This post covers it all: http://www.osnews.com/permalink.php?news_id=16524& comment_id=183548

A local-to-local symlink is one that I open from my own machine that points to another file on my own machine. As the post mentions, local-to-local symlinks are on by default. No problem there.

A remote-to-local is a symlink on a network share that points to a file on my own machine. Local-to-remote is a link on my machine pointing to a network share, and remote-to-remote is a symlink on a network share pointing to another network share. All of these have security implications.

For local-to-remote, a critical system file could be replaced to point to a spoofed copy on a network share. For remote-to-local, I could wind up copying or deleting files from my own hard drive without realizing it. For remote-to-remote, I could think I'm accessing a safe server, but I'm really being redirected to an unsafe server without realizing it.

Of course, if all of these features are turned on via group policy, you have full, classic symlink support (target always resolved by the client, and Win2k and XP clients can't resolve them).

Subversion *does* support symlinks.

[Dr_Barnowl]

... although not on Windows, obviously.

Subversion clients above version 1.1 on *nix systems which support symlinks are quite happy to support them, although it's a bit of a kludge.

The SVN repository filesystem itself has no concept of the symlink, something which remarkably enough, Visual SourceSafe *does* support. The "Share" command in VSS is about the only thing I miss about it when using SVN.

I *don't* miss the period repo crashes, the inordinate sensitivity to network outage, and the propensity of VSS to allow you do do horrible things to your data without even warning you.

And of course, if anyone really finds a feature missing from SVN to be really boiling their noodle, they can always code it up themselves (programming skills / cash to pay for developers may be required.).

Hey I just noticed this,,,

[Anonymous Coward]

My windows shortcut to firefox on my NTFS drive doesn't launch firefox when I boot into linux. Linux sucks.

Re:Windows' FS / the alternate universe story

[drgonzo59]

Yes, it could never get up to 100% the stability of a fully integrated system but it could get close by forcing the hardware companies to follow strict interface guidelines. And MS does have a big enough weight behind it to force the hardware makers to accept their standards. Then consumers would soon learn to buy only the hardware "blessed" by Microsoft.

Actually I wish Linux would have the same thing, at least the hardware makers would clearly label which products work or don't work with Linux just to make it easier to shop. I don't care if such and such a device doesn't work with Linux, I just want that clearly shown so I don't have to return it after wasting a whole day trying to install it.

By the way, Sun's Solaris is also working great with hardware it was tested on, but as soon as you try to install it on a generic PC, there will be countless problems with not finding such and such a driver.

Re:Go Windows 2000!

[hcob$]

I've got mod points, but I just had to comment.

That's probably the most insightful approach to new Windows OSes that I've seen yet. The only way it'd be better is if you can convince MS to come to your school and have a launch party for something they are putting out.

When I was a student, we had MS come to our campus to have a launch party for VS.NET. Just for signing up for the presentation, you got a full copy of XP PRO, full copy of VS.NET and were entered in a chance to win one of 2 xBoxes with 5 games...(oh... and this was 2 months after the xBox Launch).

Re:Windows' FS / the alternate universe story

[DeadChobi]

Why does it have to be something the average user will want in order for it to be implemented? That's a backwards attitude to take, when you consider all the people like me who would love to play around with something like symlinks but can't. Or all the people who would like to be able to edit video or pictures on their PC and not have to spend $600 or more on software for that... wait, I guess I could just go get a Mac.

I could set up symlinks from my music and other media on a seperate partition to portions of the start menu. That's another thing that pisses me off about Windows, that 90% of the features they have in it are completely hidden and all of them default to situations which do not apply to me or actually make it harder for me to use my computer. Like the My Music and My Pictures and My Documents folders are all linked to a folder on my C drive, which is not the storage drive on my computer and default to a directory structure which I do not like. I can move the reference to the documents folder elsewhere, but the Music and Pictures folders stay right where they are, nested inside the documents folder. I've had to hack the registry to get the references changed.

It's that kind of half-assed implementation that drives people nuts. Why do they keep releasing new versions of Media Player and Internet Explorer when they can't even make an OS that's consistently useful for everyone and includes more than the run of the mill feature set?

Yeah, let's make litter outta these literati! *nt

[Mateo_LeFou]

or is that *too clever?

Re:Go Windows 2000!

[TheRaven64]

Any Computer Science student can get free copies of all MS OSes (including the datacenter editions), their dev tools, and a load of other stuff (not Office or Virtual PC for Mac though, sadly). I grabbed a copy of XP from them to play some old games I still had.

Re:Windows' FS / the alternate universe story

[I'm Don Giovanni]

"They cannot break legacy code but Apple gambled and did it, and I think it succeeded."

Apple didn't abandon Copeland for Unix, they abanodoned Copeland for NextStep, which happened to run on a bsd variant. Apple didn't give a damn about Unix. If they could have actually completed Copeland, they would have gone with that. If NextStep had been built on top of something else, Apple would have gone with that (and there was some consideration to go with the version of NextStep that ran on NT, but Apple didn't want to be dependend on MS in that way). Unix is a side-effect of the NextStep deal.

And those "moving to Mac OS X" aren't doing it to move to Unix, they're doing it to move to Mac OS X. And what makes OSX OSX isn't Unix, but Cocoa/Carbon, which "real" Mac apps are written to (meaning, the apps that normal people use; normal folks don't use "Mac" apps written against Unix).

I wish that Copeland had succeeded, so there would be at least one mainstream OS that wasn't built atop Unix or NT. More diversity in OS design advances the state of the art. You, on the other hand, are advocating that *every* mainstream OS be a Unix variant, as if Unix is the be-all and end-all, which is fallacy.

MOD PARENT UP

[Anonymous Coward]

The parent post breaks it down, and owns all of "I'm smarter than Microsoft" posts in this thread.

Re:Go Windows 2000!

[x_MeRLiN_x]

How? From the University? Wouldn't that be individual policy?

HCL

[zallus]

Microsoft already does that; it's called the Hardware Compatibility List. But the users haven't and won't learn that it's a good idea to follow, because when they see the $60 HCL-certified webcam beside the $17.95 Taiwanese-manufactured POS, they will choose the cheaper one. Every time. At least the manuals for these things tell Windows users what to do when Windows says "This driver sucks!" after installation.

Interoperability or functionality broken?

[Anonymous Coward]

Windows Vista is either suffering from an interoperability problem (despite all the noise coming out of microsoft), or a functionality problem. Linux machines can access symlinks on other linux machines. Linux machines can access files through symlinks on MacOSX machines, other Unix machines, old-timer microsoft machines, et. al. Vista breaks this. There was functionality before, that isn't there now. There was even a microsft fanboi (an idiot named Steve) who said 'oh, its a feature, and you don't really need interoperability anyway'. So is this the vaunted vista? Less functionality, more money, and new hardware to boot? Well thanks a lot.

Re:Go Windows 2000!

[TheRaven64]

The university provides the student with a number, authenticating them as a student, which allows them to download the software from Microsoft. The cost to the institution is zero, and all departments I've visited have been part of the scheme, even if they don't use Microsoft products on any courses.

Re:Windows' FS / the alternate universe story

[Foolhardy]

He asked for questions and I ask him why NT doesn't have proper vms-esque device names (dka0, etc). The question didn't go down very well. I supose that would break too much stuff.

NT does have device names like \Device\HarddiskVolume1 and \Device\USBPDO-3. Win32, OTOH, hides them all behind the \DosDevices directory, which just contains a lot of object symlinks like \DosDevices\C: -> \Device\HarddiskVolume1. Much of the ugliness that people attribute to NT is actually a part of the Win32 layer, not the underlying system.

I can't imagine them working on GPL'd stuff and having to release the code.

The GPL allows linking to non-GPL operating system libraries, which lets them avoid releasing OS code. Microsoft distributes lots of GNU GPL'd stuff with Services for UNIX, a POSIXy environment subsystem built directly on NT, alongside Win32.

Re:Windows' FS / the alternate universe story

[Foolhardy]

You are referring to POSIX I presume. Well, have you seen any native Unix code running on Windows lately? I didn't! Windows POSIX compliance is a joke, it was more of a marketing ploy to tell their client ("we even run Unix!") but in reality it is very broken.

Are you referring to only the built-in posix subsystem? If so, I agree that it's a joke. OTOH, Microsoft distributes Services for UNIX, a real UNIX subsystem for NT, not based on Win32.

That is why you have Cygwin...

SFU works much better than Cygwin in a lot of ways, including real COW fork support, no unnecessary overhead from the Win32 subsystem (like excessive process start overhead), and no common read-write shared memory section for all processes (representing an egregious security vulnerability).

Strength of Unix file system

[Tom Christiansen]

> Why this obsession with UNIX?

As Mike O'Dell once observed, while Unix is not without its flaws,
its file system is not one of these.

--tom

You have to be admin to create them, too

[Myria]

You have to be administrator to create symlinks in Vista. It's really lame.

Melissa

Re:You have to be admin to create them, too

[LanceUppercut]

Incorrect.

That's not ext3.

[SanityInAnarchy]

You do realize how that works, right? XP can mount an ext3 FS, but only as ext2. That means no journaling. That means FAT-like [in]stability.

Doesn't work.

[SanityInAnarchy]

Windows Vista now implements symlinks directly in the filesystem....API. And it does so in a way that breaks over networks and OSes, which is what TFA was about.

This means that .lnk files are actually more useful than these wannabe symlinks, because you can actually make them work on network shares, across OSes, and so on -- basically, anything that understands the .lnk file -- although they generally contain the full path, making things somewhat more difficult. But, symlinks are actually likely transparent to anything on the same machine, meaning your software follows symlinks by default, whereas it does not follow .lnk files by default -- you'd have to code that yourself.

And neither of them is actually part of the filesystem in the way that symlinks are on Unix.

So basically, Windows Vista (a version not out yet) is trying to approach where every Unix has been for at least 10 years, and failing miserably. Is anyone surprised?

Re:Windows' FS / the alternate universe story

[jt2377]

keep pearching brother! keep pearching! Lord know the kingdom of "Information want to be free" is coming!

Re:Windows' FS / the alternate universe story

[rynoski]

ahh, yes. SFU is the thing I was thinking of, but couldn't remember.

Re:Go Windows 2000!

[Anonymous Coward]

It amuses me greatly to note that from all the comments below it seems like the majority here don't know how symlinks work (including me), and yet the progression is thus:

1) Microsoft implement something so that it meets standard but doesn't go any further;
2) Apparent existing user of Windows finds this;
3) Ignoring the fact that his current OS supports that standard to the same level or worse, he/she abandons Vista decrying it's lack of support for said item that he wasn't using anyway. Because he couldn't.

I put it to those people that they weren't going to buy Vista anyway, and thus pretending to swear off it based on this total non-story is just a really stupid excuse.

Hardly the same

[SuperKendall]

Developing a language which is the composite of a number of other languages has a good point - it helps explore new forms of syntax and Java really explored teh VM realm a lot more than most languages dared to.

C# was a direct clone of Java, from the syntax to the libraries. What did it really bring to the table that ws new? We already had bytecode compilation with Java before we ever even saw C#. It was, and is, a pointless waste of man hours that could have been better spent improving upon Java - think not on the C# and Java we have today, think instead what we would ALL have is Microsoft had devoted thought and development energy to expanding the Java that was instead of having to build it from scratch first.

Re:Windows' FS / the alternate universe story

[Eivind]

remember all those times when you tried to delete the file it won't let you, saying it is in use

That, aswell as this link-mess is just a result of undergeneralization in the FS.

Under Windows, a file is a collection of bytes without precisely one name.

Under unix a file is a collection of bytes with zero or more names. The difference is significant.

Thus, in Windows, a file can't have two equally valid names. (hardlinks in unix) At best there's "shortcuts" which are sorta like symlinks, only they're implemented as a desktop-hack so they're really more like kdes .desktop-files. Each tool that want to understand shortcuts need to parse and interpret them, they can't just pretend they're files like you can in unix.

Also, this means a file cannot have zero names and still exist, which is what prevents you from deleting open files on Windows. To allow the removal of the one name *MUST* mean to remove the file too (since every file must have precisely one name), and that ain't doable since there's a process with a open filehandle to that file. (not without killing the process or invalidating its filehandle anyway)

Under unix there is no problem: You delete the file, it now has one name less than it used to have. If it used to have only a single name it now has zero names -- which is perfectly fine as far as unix is concerned. It still has one *reference* though, the open filehandle. So the actual blocks on disc aren't freed. That only happens when the *reference-count* of the file falls to zero.

The trick is, under unix the reference-count of a file is a separate thing from the names of a file. A name is a reference to a file, but it's perfectly possible (and normal) to have other kinds of references to a file too. You can try it yourself if you've got two open shells:

• echo hallo > file.txt (creates textfile with the text "hallo")
• less file.txt
• (in other shell) rm file.txt
• Verify that file is still there in less, can still be scrolled. (well, would be scrollable if it was bigger) Disc-space is not freed up.
• close less
• disc-space is now freed as the reference-count is zero.

This, frankly, has worked since literally the 70ies on unix (if not the 60ies) its mindboggling that MS *still* hasn't managed to get this rigth.

Do you know *anyone* who seriously uses a Windows-computer and *hasn't* had the "you can't delete/move/rename this file because it is in use" ? Didn't think so. Under unix, it just works. And has worked for decades.

Re:Windows' FS / the alternate universe story

[necro2607]

"Subjectively I know more and more people who ditch their Windows machines for Macs or even Linux."

Me too, actually.. almost everyone in my circle of friends had Windows machines at home when i first met them, but now a good portion of them use Macs, with many others planning to get a Mac as their next home computer.

POSIX is closer to the metal these days

[Anonymous Coward]

You're right: it wasn't *supposed* to be closer to the metal. Unfortunately, it was designed to match, roughly, a 1988 Unix system, and in the past 20 years we have come up with some better ways to do things.

For example, POSIX doesn't have anything to say about FSRefs, or any way to deal with them -- so about all they can do on the Mac is implement POSIX as a lower-level API. If they tried to write POSIX routines on top of FSRefs, either your code would work very differently on non-FSRef-based systems, or it would require a bunch of application-level #ifdefs, both of which are precisely the problem that POSIX tried to help solve.

So yes, that's the way it works if you pretend that "file io that has been standardized for ages", but if you want to take advantage of useful new features that users expect from your system, POSIX isn't always that much help.

Re:Windows' FS / the alternate universe story

[Lord Kano]

People want to sit down at their PC and be able to watch their movies, listen to their music and play their games without trawling websites for 2 hours, reading manuals.

A well designed distro will allow the first two of those things. It's going to be a long time before the last one becomes a reality.

LK