XFS to be released under the GPL

Re:What if.... by tzanger · 1999-08-11 23:10 · Score: 1

There's such thing as IP stack fingerprinting. It's what programs like nmap do to figure out what the target system is running.

If Microsoft were to copy Linux's TCP/IP stack, their fingerprint would match as well. It's not easy to change your fingerprint without changing the code to generate what it is you're fingerprinting.

Re:What if.... by divbyzero · 1999-08-11 23:10 · Score: 1

> example: how do you prove that Win2K doesn't use some modified Linux IP stack?

Point well taken, but I think you might have picked a better example... I thought the recent benchmarks showed that the core IP stack was one of the few places where NT actually performed better than Linux. Of course, 2.3 is already far along at addressing the problem, right?

But my grandest creation, as history will tell,

--
But my grandest creation, as history will tell,
Was Firefrorefiddle, the Fiend of the Fell.

Ahem! by styopa · 1999-08-11 23:11 · Score: 2

What is going on with the comments today. Everyone seems to be saying, roughly, 'What if SGI releases this under GPL and someone other than Linux uses it! That would be bad!' Excuse me!

Ok, correct me if I'm wrong but I thought the whole point of releasing something open source was so that ANYONE could use it, so long as they published the code and allowed people to modify it. The code, although in a sense given to Linux, is not just for Linux.

Sure there could be a problem of someone like MS taking the code using it and not publishing it. They would be breaking the law if they did, and if it ever got out then they would be screwed. But assuming that they DID integrate it into Windows, publishing or not, the world wins in one sense. This is because they would be using a very good filesystem. Not only that but it would make projects that allow for reading and writing of slices containing MS would be much easier. Perhaps it will help create more of a standard for file naming.

Seriously, don't b?tch about another OS using GPLed code. It is seriously uncool.

--
Disclamer - Opinion of Person

Re:Ahem! by Anonymous Coward · 1999-08-12 16:37 · Score: 0

No, we want it to be *just* for Linux. SGI gave it to *Linux*. Ripping off a gift to someone else is "seriously uncool". BSD folks can use the code and work with it as much as they want if they move to Linux. And if they don't like Linux, they can work on it, and make it do whatever they want -- and we *know* that they can, because the GPL ensures that they have the power to do so.
Re:Ahem! by davedavedave · 1999-08-11 23:20 · Score: 1

I think they meant that the code would go into Windows, and they could benefit from it without attributing it to SGI. They would not be sticking to the license, but who would know if no-one's allowed to see the Winslop source code

--
~ Artificial intelligence is no match for natural stupidity ~
Re:Ahem! by Anonymous Coward · 1999-08-12 04:07 · Score: 0

Ok, correct me if I'm wrong but I thought the whole point of releasing something open source was so that ANYONE could use it, so long as they published the code and allowed people to modify it. The code, although in a sense given to Linux, is not just for Linux.
Wrong, the GPL also forces you to release the code under the GPL. You cant release the source code under any other licence, even if it was equally free.

Re:I'm sorry, that is just wrong. by Anonymous Coward · 1999-08-12 16:16 · Score: 0

>If you want to release in a proprietary fashion - you gotta rewrite it.

Which means you aren't releasing it in a propriatory manner. You can't. You have to create a totally new product and release it.

GPL loophole by Anonymous Coward · 1999-08-12 16:20 · Score: 0

>You get the exact same situation when you deal with software under other licenses with a number of copyright holders. You can just purchase license to roblimo's and cmdrtaco's code, and reimplement xxx.h and xxx.c yourself. That's still considerably less work than rewriting from scratch.

So the GPL doesn't protect algorithms, just code.

I can write a program that will mangle code a bit, but produce a program that does the same thing. I've just bypassed the GPL.

There's a program that translates code into plain English. You can use this on crypto source, take it out of the US, and translate it back into source. That would bypass the GPL too, since it reduces the code to raw algorithms.

Re:What about *BSD? by Anonymous Coward · 1999-08-12 16:21 · Score: 0

Non-free my foot. *BSD* isn't free. "Free speech!"

FreeBSD DOES use GPL code.... by mosch · 1999-08-12 05:30 · Score: 1

in userland (gcc comes to mind) and optionally in the kernel... to quote LINT from FreeBSD 3.2

# A math emulator is mandatory if you wish to run on hardware which # does not have a floating-point processor. Pick either the original, # bogus (but freely-distributable) math emulator, or a much more # fully-featured but GPL-licensed emulator taken from Linux.

Now let's turn the flame throwers off and realize that there are more important things to argue about than licenses. After all, I think most of us use BSD and GPL software constantly. I have a FreeBSD machine, a Linux machine, and an NT machine here at work, and I use 'em all. I know I'm not alone.

Re:What about *BSD? by Anonymous Coward · 1999-08-12 16:24 · Score: 0

The libraries are under LGPL, man.

Distributing XFS as a plug-in file system separately wouldn't work for M$. You can't use it alone, without the main program. Hence, not " considered independent and separate works in themselves".

Ext3 by Anonymous Coward · 1999-08-12 05:31 · Score: 0

isn't ext3 suppose to be journaling?

Be nice to FreeBSD... by Anonymous Coward · 1999-08-12 16:26 · Score: 0

Look, they could be out coding for Windoze, which *would* be dangerous for Linux. This keeps them happy. (No offense to BSD people...)

Re:XFS for FreeBSD by Anonymous Coward · 1999-08-12 05:35 · Score: 1

You make the assumption (as many linux guys do) that BSD presents themselves as one OS. They are different OSes, with different goals, and different developers (though there is a lot of code sharing). Linux, on the other hand, presents themselves as one OS because they share a kernel, but really there are tons of Linux distros, and they all have different packaging systems, upgrade procedures, and administration tools. The 4.4BSD derived family all have very similar packaging systems, similar upgrade procedures, but they present themselves as different OSes, not one and the same like linux. Try this: Debian GNU/Linux -> Red Hat Linux -> SuSe Linux -> Stampede Linux -> Slackware Linux -> Caldera OpenLinux -> TurboLinux -> Mandrake Linux -> LinuxPPC -> MkLinux And all the rest of the distributions. And not only that, you can say Linux is ported to _____ hardware platform, but which distributions have linux ports to that platform? Not all distributions have DEC Alpha, Sparc, or PowerPC distributions. And what's this I hear about different alpha specific distrobutions which operate differently? But, for example, NetBSD runs on all of those platforms, and if you get complete CD's, you can install for any of those platforms. With the BSD family, there are three projects, each with a slightly different goal, but when you say NetBSD runs on a Sun3, you don't have to go looking for a NetBSD distribution. So 3 different OSes, or the great number of linux kernel distributions. BSD and linux are both fragmented, but in different ways.

Re:Compatible with SGI xfs partitions by Anonymous Coward · 1999-08-12 16:29 · Score: 0

Wish you could buy individual parts of SGIs for good prices from SGI. Save up a bit of money :-) throw out the Voodoo and put in some 3d iron from the SGI.

Re:What is journaling? by Anonymous Coward · 1999-08-12 16:31 · Score: 0

So what if it crashes part way through c) writing the write verification?

And how about the performance hit? There has to be a nasty one...

Re:What if.... by Yakman · 1999-08-12 05:58 · Score: 1

example: how do you prove that Win2K doesn't use some modified Linux IP stack?

Easy. Win2K is up and down like a yo-yo because of it's IP stack on the test site. :)

Then again, it could just be the 'optimisations' the MS developers put into it. ;) Security through unavailability rules.

4.4 BSD LFS does not work on FreeBSD by Anonymous Coward · 1999-08-12 06:00 · Score: 0

He's most likely referring to the BSD-LFS. The 4.4 BSD log structured filesystem is a journaling filesystem. I too have hoped that FreeBSD would fix it. But alas, I'm still waiting also.

Re:4.4 BSD LFS does not work on FreeBSD by howardjp · 1999-08-12 06:37 · Score: 1

lfs is no longer in the FreeBSD source tree. NetBSD has fixed this though, check it out.
Re:4.4 BSD LFS does not work on FreeBSD by Anonymous Coward · 1999-08-12 06:55 · Score: 0

I read somewhere on a mail thread that log structured filesystem != journaling filesystem something to do with the log itself, but they aren't the same thing from what I've read. Besides, FreeBSD is putting more effort into Soft Updates I believe, which seems to be a satisfactory substitute for journaling for most people, even though it really isn't considered production quality yet.

Re:XFS for FreeBSD by Anonymous Coward · 1999-08-12 16:39 · Score: 0

Until a propriatory closed really agressive BSD distro comes along (the type that the GPL protects against) and kills off the others...and then the BSD communitiy starves from lack of code (which isn't forced open like GPL code).

Re:um no.. by William+Tanksley · 1999-08-12 06:08 · Score: 1

Big hint: people can already sell free software. It's part of the definition. You know, 'free', as in "you can use this to help yourself in any way, so long as you don't use it to chain someone else."

Making money is certainly helping yourself.

-Billy

Good news for Linux though by Anonymous Coward · 1999-08-12 16:40 · Score: 0

We get more disgruntled BSD developers from a dying project...

Re:Good news for Linux though by Anonymous Coward · 1999-08-12 18:30 · Score: 0

I feel sorry for the BSD guys. It is horrible. For them, it is like being diagnosed with cancer. You go home and stare in the mirror. The doctor's words echo in your head, ``you've got incurable cancer''. After hearing the diagnosis, many of the BSD followers have lashed out in rage. They are like cancer victims mad at the world for their own malady. Doctors will tell us that they are in ``denial'', the inability to face the reality of their own mortality. BSD is dying and it isn't pretty. I'm sorry for those guys not only because of their immediate problem, but because their lives are so empty that they have wrapped their souls and self esteem in the shroud of a doomed operating system. Some may get over it, but already we see many deeply disturbed individuals, many disillusioned, deeply disturbed BSD refugees.
Re:Good news for Linux though by Anonymous Coward · 1999-08-14 11:41 · Score: 0

``BSD Dead---film at 11...'' Hasn't one group or another been saying this for the last 20 years? Please, post some statistics to back this up. Someone else already posted saying that BSD is growing faster than Linux.

Re:What about *BSD? by William+Tanksley · 1999-08-12 06:13 · Score: 1

I'm glad we have free licenses like *BSD to go along with the pragmatic licenses like GPL.

-Billy

Re:Which begs the question... by orcrist · 1999-08-12 06:19 · Score: 1

To "beg the question" means to assume the thing that you're trying to prove

I thought it meant to assume that something you can prove logically entails the thing you're trying to prove; i.e.:

1. OS A has threading and OS B doesn't...

3. Therefore OS A is a faster than OS B.

Which begs the question:
2. Is an OS with threading faster than one without? (note: this is hypothetical; I don't even know the answer, and it's probably a 'depends' anyways).

Correct me if I'm wrong or if that's what you were trying to say and I misunderstood you :)

chris

--
San Francisco values: compassion, tolerance, respect, intelligence

What are you talking about?? by darkglobe · 1999-08-12 06:25 · Score: 1

My RS6000 uses regular UW scsi connectors, nothing proprietary there! However, I don't like the box either. Here's why:

1: Overpriced.
- Need I say more?

2: Poor design.
- There is NO ROOM inside the damn thing. Granted the tower has some room, the desktop version has no room, looks like a rats nest under the hood.
- Also, the sliding door on the front conveniently covers the air slots for the two UW hard disks (in our system) which causes most of them to cook. (and crash).
- Another thing, the box was DOA. This customized server (read 100,000.00$+ box) would just display some useless number on the front, and no boot/video. I called IBM (which was a battle to just talk to them) said to remove ALL the connectors from the system and reconnect. If this didn't work they would send a tech with a new MOBO.

I dismantled it three times, and no luck. The fourth time, I walked by it and in discust, I smacked it's power switch for the last time, it came up to my surprise and has been fine ever since.

Back on topic though, AIX is SLOW!! One nice thing though, it is consistently slow, 1 or 1000 users, it always works, and JFS is quite kewl! I have no problems pressing the power switch on a fully operational system, knowing that it will come back without an argument is GREAT!

Lets hope XFS can do the same for linux!

Is there a chance to get XFS for FreeBSD, too? by Anonymous Coward · 1999-08-12 06:27 · Score: 0

I would like to see XFS on FreeBSD, too. Not only because it's good, but also because of the interoperability with Linux, then. I think it should be at least possible to make XFS an option in FreeBSD (just as the GPL'd coprocessor emulation). Am I right?

Re:Is there a chance to get XFS for FreeBSD, too? by Anonymous Coward · 1999-08-12 11:51 · Score: 0
Why is freebsd dying? Each week brings more and more bad news for the dwindling number of its fans. Despite the mournful wails of the remaining BSD zealots, the leading industry analysts have coolly concluded that BSD has become a dead end as its user base continues to shrink, and ISVs abandon support. What brought about BSD's headlong spiral of death? What lessons can be learned? Here are some of the reasons put forth for freebsd's failure:
- closed development (real contributions limited to old-boy clique).
- contempt for others in non-freebsd hacker community.
- unrepentant hypocrisy (hatred of the FSF while slurping up all the FSF GNU tools in sight).
- rabid jealousy of the success of others (e.g. the success of Linux and BeOS).
- fratricidal incompatible fragmentation of BSD community (freebsd, 386bsd, openbsd, netbsd, bsdi).
- aging user base (gray beard syndrome).
- lack of true innovation (emphasis on trying to match feature check lists).
- single source vendor trap.
- devoid of third party and ISV support.
- death of CSRG and government subsidies.
Freebsd is a classic study in office politics and backroom backstabbing. The freebsd ``core team'' has been devastated in recent months, punctuated by the resignation of many of its senior members. From lax attention to security as highlighted in Usenet's comp.risks, to continued financial woes of a certain vendor of freebsd cdroms, the ongoing decline of freebsd should serve well to instruct others in how not to run a software project.
Re:Is there a chance to get XFS for FreeBSD, too? by Anonymous Coward · 1999-08-12 12:24 · Score: 0

> closed development

and by no coincidence, better code

> unrepentant hypocrisy (hatred of the FSF while
> slurping up all the FSF GNU tools in sight).

> rabid jealousy of the success of others (e.g.
> the success of Linux and BeOS).

As we all know from reading slashdot, the majority of comments come from a small minority of the users. Some BSD users hate the FSF and/or are jealous of the success of Linux and BeOS. Most don't.

> fragmentation

The BSD's run eachother's binaries, are compatable at the source level, and share code. It's really not as big a problem as you imagine it.

> aging user base

Yeah, all the cool kids use linux, what would people say if I suddenly started using BSD?

> lack of true innovation

Could you give me some examples of true innovation Linux?

> single source vendor trap

This has it's advantages and disadvantages. On one hand there isn't nearly as much competition, however you don't have stuff like "Code Warrior for RedHat Linux."

> devoid of third party and ISV support

Linux was in the same situation before it became the latest fad. Who's to say BSD won't be in the same position 5 years from now.

> death of CSRG and government subsidies

well, you can't win 'em all.

In closing, you have to remember that there's a difference between user base and market share. BSD's market share is shrinking. No one denies that. However BSD's userbase is still growing. It may not being growing as fast as other operating systems, but it's still growing. Expect to see a real jump in BSD users in a few years, as all the people using Linux to be different start to jump ship.

XFS for FreeBSD by howardjp · 1999-08-11 23:12 · Score: 2

It would be great to see the software dual licensed under an unencumbered license so that FreeBSD could include it.

Re:XFS for FreeBSD by Anonymous Coward · 1999-08-12 00:19 · Score: 0

It wouldn't be worth the effort. FreeBSD is slowly dying. It is silly to expend any effort on porting XFS to a project that won't be around in a couple of years.
Re:XFS for FreeBSD by flayum · 1999-08-12 02:14 · Score: 1

If I were a company releasing source to the public, I would probably not release it under the FreeBSD license because I wouldn't want my competitors to be able to profit from selling it. However, if my goal was simply to get the most widespread adoption of my software, I would likely do a dual license release.
Re:XFS for FreeBSD by howardjp · 1999-08-12 01:22 · Score: 1

FreeBSD is not dying. FreeBSD is growing at a faster rate than Linux is (albeit starting from a smaller user base). It would require some devistating act to kill a project that well managed and controlled. Five years from now, FreeBSD will still be around in a recognizable form and Linux will have split into several different OSes as each vendor tries to control the market.
Re:XFS for FreeBSD by Anonymous Coward · 1999-08-12 04:55 · Score: 0

hmm lets see.. BSD -> OpenBSD, NetBSD, FreeBSD OpenBSD -> NetBSD -> FreeBSD -> Linux -> Linux Now who's talking about fragmentation ?

Re:What if.... by Anonymous Coward · 1999-08-11 23:13 · Score: 0

If someone, say a disgruntled Microsoft employee, were to clandestinly leak the fact that Windows 2000 has parts of the linux kernel in it, Microsoft would be in quite a bit of trouble. Mind you, I'm sure they wouldn't outright LOSE a lawsuit over it simply because they have so much money, but their PR people would have a hell of a time smoothing things out again...

Re:JFS is verrrrrrry cooooooool by Anonymous Coward · 1999-08-11 23:13 · Score: 0

You *CAN* get JFS on Intel today directly from IBM. Unfortunately you have to use OS/2 "Advanced" Server to get it... :(

journalling FS is good by BugMaster+ChuckyD · 1999-08-11 23:13 · Score: 1

I don't have much experience with XFS, but I do use BeOS and it uses BFS which is a 64-bit journalling FS, and based on that experience I would say that a journalling FS would be a great addition to linux. Not only do you get a stupidly huge max file limit (which doesn't mean much now but will be great when you have a couple of TB+ movies lying around on your hard drive) but the FS is extremly robust... pull the plug in the middle of a large file cp/mv and not only does the machine restart normaly (no fsck) it resumes the copy from where it was when the power went out

Be also has other FS related goodies like searchable file attributes and mime types, but thats a different story.

Re:replace ext2?(8GB limit in ext2 is B.S.) by Anonymous Coward · 1999-08-11 23:15 · Score: 0

For the record... Minor Nit: I agree with most of what you say, but the 8GB limit on ext2 is FICTION. RedHats Disk Druid has problems at >8GB, and possibly some other tools... Linux fdisk has NO such limit. Disk Druid can be used AFTER fdisk sets up the drive. I have a 20G ext2 /usr/local on a 24G drive. (Yes, I have a Tripp-Lite smart UPS) Yes, fsck sucks when it happens...

Re:Who cares? by howardjp · 1999-08-11 23:19 · Score: 1

1. I do. I use FreeBSD and a journaling FS like XFS would really make my day. 2. Why would you want to? 3. Linux users also wouldn't know how to practicaly apply computing resources if you gave them a manual. 4. Who cares? 5. No it cannot, haven't you read the BSD license?

Actually by Jordy · 1999-08-11 23:19 · Score: 2

Actually, he is right. I posted this in a comment yesterday which shows at the top of the SGI Announces New Strategy and Alliance article.

I also submitted the story yesterday but they said that they were already posting a SGI story and so they wouldn't post mine. :)

--

--
The world is neither black nor white nor good nor evil, only many shades of CowboyNeal.

Not old news, significant news by wct · 1999-08-11 23:20 · Score: 1

To all the dweebs ranting about how this is old news, the original announcement was that SGI was going to release XFS under an open source license, but did not specify GPL. As I recall, someone from SGI said at the time that the company lawyers had to take a good look at the wording of the GPL to make sure it was compliant with the company's objectives. After all this time, it looks like it passed. Considering SGI must have some pretty expensive legal eagles working for them, this is a major validation of the GPL.

That a large company like SGI has enough confidence in the GPL to commit one of their major intellectual properties to it is a major news story to me (and one superior to *ahem* certain other proprietary solutions). Maybe even more important than the XFS filesystem itself.

Programming is 1% inspiration, and 99% plagiarism.

Re:What if.... by gsaraber · 1999-08-11 23:23 · Score: 1

Ah, you're right, I should have picked something different

And like someone (a bit up) said, IP stack fingerprinting can be effective too ..

Some else mentioned that the GPL is designed to allow people to learn from your code, great idea, that's what makes OSS great stuff. But it would be "immoral" to keep your code to yourself and make a ton of money of a propriatary product after you "ripped" all the good stuff out of a GPLed software.
I guess my concern is more with the morality of propriatary software (and hardware - think routers,webtv etc.) companies.

but I'm probably over-paranoid

immoral? by fremsley · 1999-08-12 16:42 · Score: 1

did i recept that right? you think the original author -- from whose code someone realizes some basic ideas -- actually _owns_ those ideas?

how much of your money you make by doing anything do you hand over to your former school teachers?

Re:Yes it is, dumbass by Anonymous Coward · 1999-08-12 16:43 · Score: 0

You can't strip the original licence from it.

Re:No it's not, dumbass by Anonymous Coward · 1999-08-12 16:45 · Score: 0

No you couldn't. The copyright would have to be reproduced in the license. The GPL is viral, ugly, and completely incompatible with anything other than public domain software, which could be put under the GPL.

The GPL is *not* ugly, the fact that it's viral is *good* (did you learn anything from programming at *all*? Viruses are tough little guys.) And it promotes public domain software. Which is good.

You BSD guys get pretty ticked over nothing. Seriously.

Re:Anti-MS moron --NTFS *is* journaling by Anonymous Coward · 1999-08-12 16:51 · Score: 0

Stupid /. angle brace stripping. Here's an even better link, tho. Here Even ZDNet says it...NTFS is journaling

Re:The Viral Nature of GNU by Anonymous Coward · 1999-08-12 06:33 · Score: 0

> all they would have to do is release the source to the IFS. Since Microsoft has never shown any willingness to do this, I doubt they would do so just to gain access to XFS.

Re:Jorunallings FS for Linux already on the cards? by Anonymous Coward · 1999-08-12 17:04 · Score: 0

Cat's out of the bag. Sean Connery is writing a journal file system for Linux. It's called the RMS file system (Reine Marie Stuart File System).

Didn't we already know this? by empath · 1999-08-11 22:13 · Score: 1

I thought this was announced near 2 months ago. Very spiffy news, anyhow.

--
"Please don't sigh like that, maam"

Re:Didn't we already know this? by Anonymous Coward · 1999-08-11 22:21 · Score: 0

Well, technically I don't think SGI had said they were going to use the GPL. If they want it in the kernel they really have no choice though. The real news here is that they released the first bit of XFS code.
Re:Didn't we already know this? by Anonymous Coward · 1999-08-11 22:30 · Score: 0

All we knew is they were going to release it. I don't remember hearing the type of license or any exact dates when it would be released.

Now we know it's GPLed--Wa-Hoo! This is very useful. I didn't have many problems with HD corruption, but I have enough of them. This will just make those recovery times even quicker....

Thank you, SGI
Re:Didn't we already know this? by spacey · 1999-08-11 22:51 · Score: 1

I believe that the GPL part of this is absolutely new. The neato part is that if this becomes an easy-to-integrate standard component, and Veritas comes out with their volume manager tools (though it would be cooler if they used the linux volume manager), then we have stock linux systems that can fail a drive, resize partitions, add disks hot (all you need is a backplane and scsi driver that lets you do this) defragment, do filesystems snapshots and other cool stuff!

Ahh! Linux as the core enterprise OS!

-Peter
-Peter

--
== Just my opinion(s)

Re:I'm sorry, that is just wrong. by howardjp · 1999-08-12 06:42 · Score: 1

I have read the GPL which is exactly why I'd never inflict it on someone else.

And that is exactly why anything I write is under the non-discriminatory BSD license.

reiserfs? by Anonymous Coward · 1999-08-11 22:14 · Score: 0

Isn't reiserfs also under the GPL? Which one is going to make it into the next major rev of the kernel? Is it really a good idea to support two virtually identical filesystem standards in Linux?

Re:reiserfs? by Mawbid · 1999-08-12 00:19 · Score: 1

Reiserfs doesn't have journalling *now*, but Chris Mason is working on it. Remember that reiserfs hasn't been released. Its feature set isn't set in stone.
By the way, Reiser has bigger ambitions than just producing a faster fileystem. He aims to change the way people use filesystems. See here for some insights.
PS. Hmm, interesting. The slashdot code seems to have removed the spaces in the anchor part of the url. Just jump to "Why Aggregate Small Objects at the File System Level?".
--

--
Fuck the system? Nah, you might catch something.
Re:reiserfs? by Anonymous Coward · 1999-08-11 22:29 · Score: 0

reiserfs is no where close to identical to XFS. reiserfs does not have journalling. As far as I know it's more of a playground for small file optimizations. Linux already has an ungodly number of filesystems in the kernel so there's no reason not to include both and have one big filesystem party. XFS is much closer to Stephen Tweedie's journalling fs effort (ext3?) or the Netware filesystem. There doesn't seem to be a lot of noise about the NWFS effort, but the Linux integration is _much_ farther along than XFS. It looks like it may be the first journalling filesystem for Linux in fact (truth?). Does anyone know if XFS has any significant advantages over Netware's FS? SMP scaling maybe?
Re:reiserfs? by John+Allsup · 1999-08-12 01:43 · Score: 1

Besides -- what's wrong with a little competition between GPL'd projects??

I also read somewhere that someone is concerned about license compatibility with FreeBSD. A better solution than dual licensing would be for the FreeBSD and Linux kernel hackers to get together and sort out a decent cross platform 'Installable FileSystem' API so that the same code can be shared across OS's (And not necessarily linked to the kernel in order to work)

John

--
John_Chalisque
Re:reiserfs? by TheProteus · 1999-08-11 23:02 · Score: 1

Seems to me that it couldn't hurt. It definately would give people a choice on which they preferred. I remember back when xiafs was an option to ext2. Of course, it's not used much anymore, but the code is still out there, and people have the option to use it if they wish.

If it comes from man, it will fail.
If it comes from god, It will succeed.

--
Detachment 3 Media
Exposed, Exploited, Exploded

Some linux distros use BSD code . . . by Anonymous Coward · 1999-08-12 06:42 · Score: 0

I remember booting an old version of slackware (3 years ago maybe?) when I just was trying to install linux on an old 386 and during the boot I see "Copyright Regents of the University of California . . ." I don't know what its like now, but I wouldn't be surprised if there was still BSD code in linux distros. I agree that there are more important things than licenses to argue about. But really, why do people always have problems w/ BSD license, but they think nothing of using X?

Re:Anti-MS moron --NTFS *is* journaling by howardjp · 1999-08-12 19:14 · Score: 1

Sorry, I knew that. I just cut and pasted that list of supported file systems.

Old News!!! by The+Dodger · 1999-08-11 22:20 · Score: 1

First Hemos, now Neal.
How many more postings must we put up with telling us "news" which has already been mentioned and covered in previous posts and comments?
Malda - it's time to exert some editorial quality control over /.
D.
..is for Defunct.

Re:Old News!!! by GnuGrendel · 1999-08-11 23:30 · Score: 1

Well, I saw it on Tuesday, and I assumed everyone else did too, in this article right here on slashdot. Check the comments... One of the score=2 (the highest score of any of the comments in that posting) is entitled "XFS is being released GPL"

Maybe you should read /.
Re:Old News!!! by Zack · 1999-08-12 00:54 · Score: 1

My apologies for sounding snippy...

However, I still don't think that posting something in a comment yesterday makes this old news today. I don't read every comment on every article on a slashdot... I read a lot of them, but I don't always have time to read them all. I think that the fact that XFS is being released under GPL is enough for it's own article.

And I also hate it when people who whine about slashdot. If you don't like it, don't come back.
Re:Old News!!! by Zack · 1999-08-11 22:29 · Score: 1

Old News? Oh really.. when did they tell _you_ how they were going to release it? That's right.. you know MONTHS ago if they were going to use a BSD style liscense or go with the GPL. Well it's very rude of you to not tell us. In case you're too thick to get it, this IS news. We knew XFS was being released but not how. Now we know both.

>How many more postings must we put up with
>telling us "news" which has already been
>mentioned and covered in previous posts and
>comments?

None. Stop reading this site you fscking whiner.
Re:Old News!!! by The+Dodger · 1999-08-12 15:57 · Score: 1

OI! YOU! OFF MY PLANET!
You suck, you're a lamer and you're a fuckwit.
However, I still don't think that posting something in a comment yesterday makes this old news today.
Well, you're wrong.
I don't read every comment on every article on a slashdot...
Well, you should.
And I also hate it when people who whine about slashdot.
I'm not whining. I'm in FRM (full rant mode).
If you don't like it, don't come back.
I'm willing to bet that Rob would rather know about it when people are getting pissed off at something about Slashdot, rather than have his traffic slump for no apparent reason.
And never apologise. It's as good as admitting that you're wrong. Oh wait, you are. Forget it...
D.
..is for Dickhead.

That would be illegal, dumbass. by howardjp · 1999-08-12 19:23 · Score: 1

As I said to some other idiot, the BSD code cannot be relicensed under the GPL. That is just as illegal as relicensing GPL code under a BSD license.

Re:I'm sorry, that is just wrong. by howardjp · 1999-08-12 06:50 · Score: 1

The fatal flaw with your counter example is that if I wanted the features of package Z, I'd have put them into package Y to begin with :)

Which begs the question... by Snack+Cake · 1999-08-11 22:20 · Score: 1

When can I get it?

Re:Which begs the question... by Nat+Lanza · 1999-08-12 09:39 · Score: 1

I think part of the confusion is that most people don't realize that "beg" has a meaning other than "ask for". It can also mean to evade or sidestep. That's the usage in "to beg the question" -- "begging the question" means "avoiding the matter at hand", not "asking for a different question".
Re:Which begs the question... by Nat+Lanza · 1999-08-12 00:01 · Score: 1

It doesn't beg any questions at all. To "beg the question" means to assume the thing that you're trying to prove. The common usage of it as "leads to the question of" is completely wrong.

Sorry, it's a pet peeve.
Re:Which begs the question... by Snack+Cake · 1999-08-12 02:36 · Score: 1

Ok. Good point.

Oh great... by Anonymous Coward · 1999-08-12 06:53 · Score: 0

Just terrific. Now, if I need to implement something like it, I'll have to re-engineer it to get around being trapped by the infamous General Public Virus.

Re:Freebsd is dying by howardjp · 1999-08-12 19:36 · Score: 1

closed development: This is simply not true. Last month I suddenly turned up on the mailing lists posting code and it has all been welcomed by the community.

contempt: This is rare.

hypocrisy: There is current work being done replacing those GNU utilities which are present. Anyway, FSF is just as guilty.

fragmentation: Free, Net, Open, and BSDi are not the same OS, do not characterize them as the same. 386BSD has been dead for nearly a decade.

aging user base: I started using BSD three years ago when I was 16.

innovation: CAM, UVM, and this is only in the past couple months.

single source: Anyone can resell BSD (Cheap Bytes, Walnut Creek)

third party: The ports tree (a place where you can install any third party program with a FreeBSD port with one statement) contains nearly 3000 applications.

CSRG: That was 5 years ago, since then, BSD development has done nothing but grow. BSD is not dying.

replace ext2? by coreman · 1999-08-11 22:20 · Score: 1

I can't see it replacing ext2 for the general user distros. I can see value if you're running a server site or some large raid DB but for day to day stuff.

Re:replace ext2? by Anonymous Coward · 1999-08-12 01:28 · Score: 0

Besides no more fsck, XFS is ~10% faster than SGI's previous FS which was no slouch either. I confirmed SGI's claims to this effect with my own primitive tests and also verified that it speeds up I/O-intensive apps by up to 10%.

Not bad in my book.
Re:replace ext2? by otis+wildflower · 1999-08-12 03:37 · Score: 1

CALM DOWN EVERYONE!!!!

There's no reason why you couldn't continue to use ext2 filesystems for whatever reason you'd want, until the ext2 drivers are excised from the kernel source. However, having the ability to boot from a mirrored logical volume (plex? my AIX -> Veritas filter is faulty) without paying OS/Veritas tax is extremely appealing to those of us who care about data integrity, availability, etc...

Flexibility and features are GOOD, particularly if they're free and robust....

You don't know jack shit, retard by howardjp · 1999-08-12 19:41 · Score: 1

Dumbass, that would be illegal.

Re:You don't know jack shit, retard by Anonymous Coward · 1999-08-14 11:37 · Score: 0

Isn't that what he said, bung munch?

It's illegal, dumbass by howardjp · 1999-08-12 06:54 · Score: 1

That would violate the license of the Regents of the University of California, just as putting a BSD licensed on GCC would violate the license of the FSF.

sounds cool...but by bendawg · 1999-08-11 22:23 · Score: 1

So does anybody have an idea how long it will be
until XFS is actually supported by the kernel? Hopefully soon.

Re:considering... by chuck_g · 1999-08-12 07:01 · Score: 1

No, what he said is that a journing fs would probably not make 2.4.0. We may see it in a patch release later and hopefully before 2.6.0.

A pre-emptive strike by Wohali · 1999-08-11 22:25 · Score: 1

...most likely against IBM, who is poised to port their journalled file system (JFS, which rox my world) to their "project Monterey" version of a semi-free, semi-commercial x86 UNIX.

Has anyone ever run comparisons between JFS and XFS? I'd be interested in reading the results of a head-to-head comeptition. I've always been a fan of JFS on my AIX boxen. There's nothing like experiencing a hard power failure and having your UNIX box come back up without so much as batting an eyelash.

--
"But always she's the spectre of uncertainty I first endured, then faded, then embraced..."

Re:A pre-emptive strike by Precision · 1999-08-12 00:09 · Score: 1

Just in case you didn't know.. XFS and JFS are the same filesystem. They both are written by and licensed from VERITAS Software (VXFS). It is also the same filesystem that goes into many other commercial unices. Just thought I'd let ya know.

--
- U
Re:A pre-emptive strike by Anonymous Coward · 1999-08-12 01:11 · Score: 0

I'd have to give a big no to the XFS from veritas. You are thinking of the replacement for Sun's UFS, NT NTFS, etc. XFS is even better than Veritas in some features that nobody has mentioned here yet. An extermely good feature is that there is no penalty for having lots of files within a directory. Normally I/O starts to seriously degrade when you have 100+ files in a single directory, XFS you don't have that problem. Of course Veritas has some nice features that XFS doesn't have currently (HA and some others), like they say all software sucks, some just suck more.

I talked to a upper echelon Veritas guy at Usenix, and he mentioned that SGI & Veritas have had some very bad blood in the past. I had been trying to find a HA solution for Irix, and he said that Veritas for Irix won't be available any time soon. Of course I think I saw a mention that SGI & Veritas were working together on a product a month or so ago, so they may be friends again.

tsuiter@midusa.net
Re:A pre-emptive strike by Anonymous Coward · 1999-08-12 15:30 · Score: 0

How the h*ll do you run comparisons between filesystems? You testing their drivers, OS disk I/O subsystems, or drive rates?

You gonna spit on the drives and see which ends up with more corrupted data?

Re:Guaranteed I/O rate on XFS by mtest · 1999-08-12 07:33 · Score: 1

Actually, according to the IRIX manual for grio, it has much to do with the OS; it will have to do resource evaluation in order to know if the hardware can cope with the request, then some resource reservation stuff; the whole thing is designed for any kind of resource reservation, but it only works on xfs/real-time partitions.
I suppose this is the reason why they put that disclaimer in the XFS/Linux presentation: it might be too much trouble to introduce this in the linux kernel.

considering... by davedavedave · 1999-08-11 22:27 · Score: 1

that Linus recently announced that the journaling fs would have to wait till version 3, this is good news

--
~ Artificial intelligence is no match for natural stupidity ~

The Viral Nature of GNU by sterno · 1999-08-11 23:26 · Score: 1

GPL's got a very nice built in safety to keep that from happening. If you are a company that bases your business model on closed-source software, you take a very very big risk by including GPL'd code. If you are caught, then all of your source code will be exposed to the world because it was based on GPL code.

So if MS decided that XFS was better than NTFS and decided to coopt it and make a new filesystem, they would risk the entire source code of NT in doing so. It's not so much the likelyhood of being caught that keeps them in line, but the severity of the consequences if they are caught.

---

--
This sig has been temporarily disconnected or is no longer in service

Re:The Viral Nature of GNU by Anonymous Coward · 1999-08-12 15:50 · Score: 0

BS. They wouldn't have to give up their code. They'd pay a fine or something and remove the code. They wouldn't open up source to Windows if some disgruntled employee manages to slip some GPL code into it.

You really think you could read and understand the Windows source?

Why would you read it, anyway? What are you going to get out of it?
Re:The Viral Nature of GNU by Anonymous Coward · 1999-08-12 01:54 · Score: 0

if microsoft did in fact incorporate the GPL'd XFS code into a new version of NTFS, and were caught (as you have speculated), wouldn't they just have to release the source to the NTFS component of the NT system? or would they truly have to release the entire NT source? my point is that since they only ripped off a the XFS code for filesystems, and only incorporated it into a part of their OS (the filesystem part, duh), wouldn't they only have to release the filesystem source and not stuff like the GUI, kernel, device/IO control, etc.? I've never actually read the *entire* GPL, so i am by no means an expert in this area. can someone help me out & answer my question?

thanks.

p.s. something to illustrate my point: someone a little farther down wrote (under the subject, "The real reason...") that SGI had ripped off some GPL'd code for XFS. you don't see them opening all of IRIX now, do you? just the XFS code. (if this claim is true, then it serves to illustrate my point, but if not, just disregard the p.s.)
Re:The Viral Nature of GNU by Mouse · 1999-08-12 04:23 · Score: 1

Microsoft could easily include XFS lock, stock, and barrel without a fear of releasing the NT source because NT utilizies installable file systems(IFS). Therefore, the filesystem code is external to the kernel source code. As such, all they would have to do is release the source to the IFS.

-Mouse
Re:The Viral Nature of GNU by Anonymous Coward · 1999-08-12 16:00 · Score: 0

If the external file system is essentially external, with only a few secret APIs at stake, if XFS became a selling point, you bet they'd do that (or put up FUD about NTFS being better -- it *does* journal, after all).

Re:What is journaling? by UnknownSoldier · 1999-08-11 23:41 · Score: 1

The problem: When a Operating System crashes, the File System is in an unknown and potentially inconsistent state (yes, a BAD thing ;)

The Solution: Basically keeping a log of all file/disk actions, so we can "restore" and put the system in a known consistent state.

i.e. A user saves a file in an application
The File System needs to:
a) access the meta-file area, and create a new entry (i.e create the file)
b) allocating space for the file and write the data (commonaly called commiting the data.)

If the Operating System crashed during step b,
the next time the user tried to read the file back, it would contain garbage since the meta-data said it was a certain size, but not all of the data was written.

With journalling, i.e. there would be a step c) Log the file write as complete.

When the system restarts, the File System would start at the last logged action and work its way to the first logged action undoing all the steps, putting the file system in a known and consistent state.

Usually the log is implemented in a circular buffer, since keeping logs of EVERY event since startup would quickly fill up disk/memory space.

Not the best explaination, but it should help.

Re:Who cares? by vulcan · 1999-08-11 23:46 · Score: 1

So it seems the point of the GPL is to prevent people from using the code. Sounds an awful lot like proprietary software licenses to me....

sc

Question from the uneducated by ashpool7 · 1999-08-11 23:47 · Score: 1

Does this mean any changes to the partitoning system that linux uses? Do all those ex2fs partitons in my extended just get converted to XFS if this is implemented? Or does a "new" scheme arise...

[offtopic]
Personally, I would appreciate this upgrade if a "new" scheme were implemented, especially if it was similar to the FreeBSD or Solaris way of handling multiple filesystems. Does nobody else think that it's silly that Linux filesystems take up more than 1 logical partiton on your drive? (I speak from the IDE/PC realm) It's a pain in the ass, IMHO, for me to encapsulate 4 to 6 Linux partitons inside an extended partiton just so I can fit 3 other "normal" partitons on the drive that don't require a boot manager to boot. But of course, to even boot Linux in that extended partiton, I HAVE to use a boot manager to boot Linux.

Wouldn't it be nice to have 6 Linux filesystems sit inside *just one* Linux/XFS logical partiton? Then you can install it with 3 other OSes and just switch the active partiton when you want to boot somewhere. No boot manager needed!

I'm probably missing something in my logic. Feel free to tell me if I am!
[/offtopic]

Re:Question from the uneducated by Anonymous Coward · 1999-08-11 23:53 · Score: 0

You want it? Then you write it. Who the hell is stopping you? No one is holding a gun to your head. So why don't you just get off your lazy ass and do it, if that's what you want. Then after you you have debugged it and it works. Tell us about it. There's a slim chance that someone might actually be interested.
Re:Question from the uneducated by McAlister · 1999-08-12 09:12 · Score: 1

The problem that you are talking about is not a Linux problem, it's a problem with the Intel/IDE BIOS "Standard", which specifies that IDE drives may contain no more than 4 "primary" partitions. AFAIK, SCSI is un-encumbered by this restriction, which is why none of the other UNIces suffer from this apparent "bug". (they all run on hardware where SCSI is the standard way of doing Disk Storage)

I'm not really qualified to comment on the boot manager issue, but it begs the question, why wouldn't you want to use a boot manager, it makes life so much easier than toggling active partitions in fdisk.

McA
Re:Question from the uneducated by Anonymous Coward · 1999-08-12 01:25 · Score: 0

First off, there's absolutely no reason to be an asshole.

Second off, as far as "a slim chance that someone might actually be interested", you're incorrect. Personally, I like the way BSD does its disklabeling. I'm uncertain of the effects (good or bad) on performance, stability and flexibility, but I've the opinion it leads to a cleaner setup (especially if you've multiple OS's each with multiple filesystems).

BTW: the kernel already has support for BSD disklabeling in order to mount/read (experimental write I believe) UFS partitions.
Re:Question from the uneducated by howardjp · 1999-08-12 01:30 · Score: 1

You know, FreeBSD will let you maintain multiple partitions within a slice.

Good/Bad? by Spyky · 1999-08-11 23:49 · Score: 1

First, this is great news because in addition to adding enterprise level (pardon my buzzwords) features to Linux, it also shows SGI's commitment to Linux. Hopefully they will continue to release some of their code (and develop new code) under the GPL.

I wonder though, at the intent to REPLACE the ext2 filesystem for linux. Yes, I agree that having a journaling filesystem is great for large Linux fileservers, webservers and database servers. Slashdot would certainly benefit from such a system ;-) However, I wonder at the performance hit caused by logging. If I remember correctly, most logging systems work by writing a message to a "log" before changing the filesystem and another after the change is complete. This would greatly increase the disk usage overhead on a desktop (single disk) Linux system like mine. Having such a robust filesystem isn't NECESSARY on my desktop. Is there a way to turn logging off (and remove the performance hit) when using this filesytem. Otherwise I will stick to ext2. Any thoughts?

Spyky

Re:Good/Bad? by Eidolon · 1999-08-12 00:03 · Score: 1

It's not clear whether you are worried about running out of disk space or feeling a "performance hit". Linux is pretty fast (one of the many reasons we all use it). If you are doing desktop-oriented things with your desktop system, you might not even notice a performance hit.

Drives are so cheap now you might not need to worry about logs taking up lots of space, either.

It would probably be worth investigating journaling file systems even if your primary use is fairly userly. ext2fs is fast, but it is also one of the most fragile file systems I have ever encountered. IMHO, ext2fs is one of the things about Linux that is mostly likely to deter people from using it in high-end servers. I have lost entire ext2fs file systems irrevocably after power outages or developer-kernel crashes. It's good news indeed that more alternatives are on the way.
Re:Good/Bad? by Brian+Knotts · 1999-08-12 10:06 · Score: 2

My only gripe about ext2 is its tendency to fragment. It's nearly as bad as NTFS in this regard. ufs and HPFS do a much better job at resisting fragmentation.
So, does anyone know if XFS is better at resisting fragmentation than ext2?

--
Interested in XFMail? New XFMail home page
Re:Good/Bad? by Spyky · 1999-08-12 00:35 · Score: 1

I am referring the performance hit caused by requiring 3 writes instead of one to modify any part of the filesystem. One of the major problems, is that the log is usually stored in some physical part of the disk that is not necessarily near the data being written, requiring the head to travel across the disk twice for just one write. In large file servers this can be fixed by hardware means, having multiple disks, etc. But in a desktop this is not the case. You may be right about "not noticing" the hit while doing standard desktop operations (browsing, WPing, etc.), I don't really know how much of a problem it causes, or if it will be significant in this case, its just a though... :-) As for the problems with ext2, I haven't really had any problems. Granted I don't do mission critical stuff on my linux machine, yet... Then again I don't do mission critical stuff :-) I don't think it has any more inherent flaws the HPFS or NTFS.

Spyky

Why? by IntlHarvester · 1999-08-11 23:49 · Score: 2

Why would MS (or another vendor) "steal" XFS-GPL, when they could just port it to the Windows NT IFS layer and include the source code on the CD?

MS might be evil enough to be stealing code from open source software, but if they are, it's purely out of pride, because there's legal ways they could use the code.
--

--
Business. Numbers. Money. People. Computer World.

experience with XFS and JFS by jetson123 · 1999-08-11 23:50 · Score: 3

I have been using XFS and JFS-based systems for several years and I'm unimpressed.

I have more experience with JFS because I have used it in some large applications; I have been running XFS on my desktop Irix machine for a couple of years.

Here are some observations about JFS; most of those also apply to XFS:

It can be quite slow in practice; for example, untarring a large amount of scientific data, JFS took three times as much time as ext2, even though it was running on a faster SCSI disk. I haven't done XFS measurements, but XFS also feels sluggish in practice to me.
It only protects file system structure, not file content.
It only protects against a small set of failures. For example, hardware failures and file-system related bugs still cause data loss.
JFS comes with a LVM (volume manager) and XFS integrates XLV. In my experience, those kinds of systems complicate disk management, increase the risk of file system management mistakes, and make it more difficult to predict performance.
Journalling does not guarantee fast recovery. There may still be extensive recovery going on at boot time. IBM's JFS, ironically, often runs on systems that have (unrelated) performance bottlenecks in their boot code that makes them some of the slowest booting UNIX machines in existence.
The only time I have actually lost a partition over the last decade was on a journalling file system due to, what appears to have been a software bug in the fs code. Journalling file systems are tricky pieces of software to write.

Of course, with XFS on Linux, we can finally compare these issues side-by-side on identical hardware and kernels. It will be interesting how XFS holds up.

XFS has some nice features, and I think it will be a great addition to Linux as an optional file system. Its availability will make Linux much more attractive to some corporate buyers.

But before adopting it widely, I believe the issues that I raise above need to be looked into and evaluated carefully. I suspect many people don't need the features of XFS, overestimate the safety of journalling file systems, and will get bitten by the complexities and overhead. I hope XFS will not preempt the further development of "traditional" Linux file systems.

Re:experience with XFS and JFS by ratkins · 1999-08-12 09:13 · Score: 1

Just to make sure we're all talking about the same thing here, I'm referring to IBM's jfs, which is part of AIX. I've been using AIX for the last 3 years or thereabouts, mostly AIX version 4 onward. I've got to say that some of your criticisms are unjust. Firstly, yes, the jfs only journals the structure of the filesystem, not the data, but what do you want for a 4Mb journal log?
Secondly, I have never had any corrupt data on a JFS volume. It's quite neat to be able to hit the power at any given time and not frag the machine. The AIX LVM also is an absolute lifesaver. Yes, it's complex to understand initially (the LVM takes about a day for me to teach in IBM's sysadmin course) but once you know what you're doing it makes life sooo much simpler. Run out of space? Just extend the filesystem. Run out of disk? No problem, just add another one to the volume group (while the machine's still on if you've got hot-swap SSA or SCSI).
I really couldn't live without it.
Re:experience with XFS and JFS by spacey · 1999-08-12 12:23 · Score: 1

1) xfs and jfs are different beasts. There are, I believe, 2 jfs' also. A Journeling filesystem, I believe, usually refers to using a log for metadata only. A log-structured filesystem uses logs for both data and metadata.

2) WRT write times on a journaled files system:
if you set your ext2 filesystem to running in synchronous mode, then you may be comparing apples to apples. An ext2 fs is, be defult, not synchronizing to dis, but to cache.

3) Re: fast recovery: Using a log-structured filesystem does guarantee a recovery time of less then the time needed w/o it. Because if data and metadata is written to a log and the system goes out, the OS is guaranteed to never have to evaluate more of the system status then what is contained within the log(s). As I mentioned before, JFS may only journal metadata.

4) Log-structured filesystems are extremely fast for small-file queues. Also, log-structured filesystems can easily have features built upon them like defragmentation, snapshots, resizing, etc. because there is no longer any dependancy on the physical aspects of the disk.

5) Using VxFS w/ solaris, I've never had a critical bug, and I've been able to add space to a volume used by oracle while oracle was running to add tablespace to a production database. That shit's cool. And xfs should be able to do that w/o any problems.

-Peter

--
== Just my opinion(s)
Re:experience with XFS and JFS by afkmn · 1999-08-12 01:06 · Score: 1

It can be quite slow in practice; for example, untarring a large amount of scientific data, JFS took three times as much time as ext2, even though it was running on a faster SCSI disk. I haven't done XFS measurements, but XFS also feels sluggish in practice to me.
It will definitely be slower. It's effectively amortizing the fsck over the write time of the filesystem. Not as slow as lfs, but slow. Note though that ext2fs is an extremely fast filesystem partly due to its fast-and-loose default behavior wrt writes. ufs, by default, writes every byte through to disk. ext2fs just writes to cache.
It only protects file system structure, not file content.
Which is why it's faster than lfs...
It only protects against a small set of failures. For example, hardware failures and file-system related bugs still cause data loss.
But that set (abnormal terminations, improper shutdown...) constitutes the vast majority of errors in practice.
JFS comes with a LVM (volume manager) and XFS integrates XLV. In my experience, those kinds of systems complicate disk management, increase the risk of file system management mistakes, and make it more difficult to predict performance.
complicate perhaps, but it's pretty convenient. Most of my experience is with advfs under Digital Unix, and it was a godsend. Want to see what usage will be like before you divide a new disk into partitions? Make filesets and put hard limits later. Need more space on a low-priority partition? Stick another disk in. Want to reduce restore-from-backup headaches? Make a nightly snapshot. It's not perfect, and there are tradeoffs, but it's nice to have around.
Journalling does not guarantee fast recovery. There may still be extensive recovery going on at boot time.
There are no guarantees in life. But our experience has been that real repair has to happen maybe once every 20 bad reboots, if that. YMMV.
The only time I have actually lost a partition over the last decade was on a journalling file system due to, what appears to have been a software bug in the fs code. Journalling file systems are tricky pieces of software to write. Of course, with XFS on Linux, we can finally compare these issues side-by-side on identical hardware and kernels. It will be interesting how XFS holds up.
This may be a valid concern. I don't know anything about the buglessness of XFS specifically.
XFS has some nice features, and I think it will be a great addition to Linux as an optional file system. Its availability will make Linux much more attractive to some corporate buyers.
There's the rub. And it's made me excited enough to try linux again after a FreeBSD vacation.
Re:experience with XFS and JFS by your+jesus · 1999-08-12 01:36 · Score: 1

Actually, I have had the opposite experience...

Working with mail servers and news servers that have massive amounts of small files (all less that 1MB, most around 4k), we see a lot of problems with ext2. When we have a systems charash (though it be rare) we must get up and running immediately.

My experience with ext2 exhibits extrememly painful fscks and information loss due to controller errors (the person bringing ir back online and walking it through the fsck).

On the other hand, XFS (no experience with JFS) is EXTEMELY quick about these things. I can bring online my 50 GB file system in minutes rather than hours.

To address the XFS is slower issue, I do a lot of I/O intensive processing and I find that ext2 is faster for somethings and slower for others, but neither is substantially better.

We need solutions to handling incoming/outgoing messages at about ~500k hour on a single machine, and we con't have the resources for solid state drives.

As we run linux on those machines, we have purchase many small (2-4GB) SCSI LVD drives and split the mail spool over all of them (14). We run ext2 and this FLY!

I personally believe that once XFS is available for Linux, I wil not have to do this kludgery and I can move back to having my spool on 1-3 disks.

Using XFS on my SGI's running IRIX has been nothing but a pleasant expereince. Using IRIX itself has been less than pleasant (not painful, though).
Re:experience with XFS and JFS by Anonymous Coward · 1999-08-12 01:52 · Score: 0

>JFS comes with a LVM (volume manager) and XFS integrates XLV. In my experience, those kinds of
systems complicate disk management, increase the risk of file system management mistakes, and make it more difficult to predict performance.

In my experience, this isn't true. AIX's logical volume manager was always simple to work with. I believe the only thing that could be accused of "complicating things" is the availability of raw volumes for apps (primarily rdbms's) that need it.
On the other hand, the ability to resize a filesystem on the fly is an _extremely_ nice feature.

You're right about one thing--having a logical volume mgr increases the risk of file system management mistakes. On the other hand, this really isn't a fair comparison. It's sort of like saying that it's hard to have an automobile accident while you're walking. In other words, logical volume managers give administrators tools and capabilities they wouldn't normally have.

>Journalling does not guarantee fast recovery. There may still be extensive recovery going on at boot time.

In general, journalling will give you a faster recovery in the event of a system crash. This is especially true if you are dealing with extremely large database systems.

In 7 years of working with AIX systems, I only _once_ needed to fsck one by hand (the root filesystem of all things).

>IBM's JFS, ironically, often runs on systems that have (unrelated) performance bottlenecks in their boot code that makes them some of the slowest booting UNIX machines in existence.

Who cares? Enterprise users tend to put their machines in datacenters with niceties like UPS, HVAC, and halon. This is the world of once yearly reboots.

>The only time I have actually lost a partition over the last decade was on a journalling file system.

The only time I've lost a filesystem ever was under ext2fs following a power outage. What does this mean about the quality of ext2fs. . .pretty much nothing as it's only one example.

I think you protesteth too much. In my experience, a journalled file system + logical volume manager is a big win for any production system.

BTW: I also hope the XFS doesn't preempt the further deveopment of "traditional" Linux file systems.
Re:experience with XFS and JFS by Anonymous Coward · 1999-08-12 03:42 · Score: 0

It can be quite slow in practice; for example, untarring a large amount of scientific data, JFS took three times as much time as ext2, even though it was running on a faster SCSI disk. I haven't done XFS measurements, but XFS also feels sluggish in practice to me.

According to this paper, XFS has "near-raw I/O performance". It also has a "guaranteed I/O rate" feature for real-time applications. SGI has more specific tests (comparing XFS and raw I/O performance) in one of the other papers on their site.

XFS uses B-trees to store directory entries. According to their graph (in the above paper) a directory with 50000 files is accessed at the same speed as a 500-file directory.

I don't know how well XFS scales to *small* file systems (ie. desktop Linux installations), but it looks like it will work very well for servers.

Re:The real question is; HOW MANY DEVELOPERS! by Anonymous Coward · 1999-08-11 23:51 · Score: 0

After cutting 1500 employees and restructuring, I don't see them having significant resources available to do this. Yeah they want a journaling filesystem on linux... but they don't want to pay for the R&D.. which is what's been killing them all this time. Porting a fully featured filesystem to linux is a pretty monumental task... even providing yer just trying to port the 20/80 stuff.

Re:bsd faces new lawsuit by howardjp · 1999-08-12 19:46 · Score: 1

A quick search of Bloomberg revield nothing. Post a link. Otherwise, these are just more anti-BSD lies.

you moron by Anonymous Coward · 1999-08-12 19:47 · Score: 0

have you read the BSD license. It pretty much says that you can redistribute with or without modifications in binary or source as long as the copyright notice stays in tact. You can't get any freer than that without it being in the public domain

Re:I don't see a problem. by Anonymous Coward · 1999-08-12 20:04 · Score: 0

I'll tell you what the problem is. I don't know why a lot of linux users don't get this. The FreeBSD community(as well as other BSD's) has a different goal than that of the FSF. The BSD community's goal behind the BSD license is that of making their code as widely used as possible. That means whoever(including MS) for whatever purpose. Whereas the FSF wants the code to be in the open(basically an anti proprietary view). That is why the FreeBSD community is not going to switch to GPL and also why they are trying to replace as much GPL'ed softare in their os as possible.

Re:Jorunallings FS for Linux already on the cards? by wangi · 1999-08-12 20:23 · Score: 1

And? I don't think 'standing on someones toes' is the issue here. If Steven Tweedie gets his journaled file system together, then fine - if not, he's missed the boat!

After all he has been at it for 'some time' now.

Choices = better for all by Cef · 1999-08-12 07:40 · Score: 1

Just because there will be more than one choice of file system for linux, you shouldn't go nuts about what might happen. Choice is nothing to sniff at, as we get exactly (or close to) what we want.

In fact, if they are all GPL (and I imagine they would be) then what you may see happen is a merging of technologies into a better product. Say one file system has some sort of shortcoming, but another system doesn't, then you might see a crossover that will result in better products all round.

I can't wait for XFS, or any other "proven" filesystem (which means that its been tested and works as intended) for Linux, as then I won't have to worry so much about power dropouts. Even with UPS's, sometimes even they fail (as happened to me on the weekend).

Re:Microsoft, read this and weep. by Anonymous Coward · 1999-08-12 07:54 · Score: 0

NTFS is not that good. Plenty of people recomend running webservers with FAT filesystems because they are faster than NTFS.

Re:Replace ext2? Yes! Ext2 has 8GB limit + slo fsc by mab · 1999-08-12 20:24 · Score: 1

Well we have a 120G RAID5 an it can take 3 hours to fsck

I can't wait for XFS

Haha by Anonymous Coward · 1999-08-13 00:48 · Score: 0

Starves from lack of code? Aggressive proprietary BSD distro? For one thing, BSD doesn't have "distributions." FreeBSD, NetBSD, OpenBSD, BSD/OS are all different OSes, but the came from a common ancestor (4.4 BSD/lite).

How could a proprietary company kill off a free project? And they won't starve from "lack of code" cuz there are still free coders out there making (revolutionary concept) NEW CODE!! Coders may not want or need what the proprietary "distrobution" brings.

And in case you didn't know, there is a proprietary BSD, BSDi sells BSD/OS, and has been around for a few years. 4.2 BSD was the basis from many commercial Unices, including SunOS. But the open source BSD family is still around.

2 GB is an x86 kernel limit by Anonymous Coward · 1999-08-12 07:55 · Score: 0

ext2 on an Alpha can have insanely huge files.
XFS on a Pentium can only have 2 GB files.

Re:2 GB is an x86 kernel limit by Anonymous Coward · 1999-08-12 15:38 · Score: 0

Gee, and my entire Linux partition is 2GB. I wish I had your problems. :-)

HFS is a drag... by Anonymous Coward · 1999-08-12 07:57 · Score: 0

... and Apple should drop support for any dual-fork filesystem in their upcoming MacOS X. MacHFS and Appletalk (the most bandwidth-inefficient protocol ever developed) are the bane of any homogenous network (though I'll admit that netatalk+asun and Helios atalk work great).

Re:I'm sorry, that is just wrong. by Anonymous Coward · 1999-08-13 00:52 · Score: 0

>If you want to release in a proprietary fashion - you gotta rewrite it.

Which is the reason for the BSD license: spend more time for everyone (including businesses) to "advance the art" instead of reinventing wheels if they want to close their source.

Re:Who cares? by davet · 1999-08-13 02:01 · Score: 1

If you wanted it fixed, why didn't you fix the code yourself? Frankly, LFS was never really a production quality filesystem, and now with SoftUpdates, I don't miss it in the least.

If people were rude to you, it's probably because you whined and didn't try to contribute.

Re:JFS .... and LVM? by CelestialScum · 1999-08-12 08:30 · Score: 1

the discussion on LinuxToday has gone on for some time about the 2.4 kernel. I don't know if the dynamic resizing will be there, but this is what was said about LVM:

Joe Drew - Subject: Re: Raw I/O, LVM (1999-07-30 05:36:48)
In fact, both of these features will be added for 2.4. LVM support and raw I/O support are both already in; LVM is undergoing quite a
bit of hacking activity, but should be ready for 2.4.

Nope, sorry... by MenTaLguY · 1999-08-13 02:12 · Score: 1

So the GPL doesn't protect algorithms, just code.

Correct. This is true of all licenses.

I can write a program that will mangle code a bit, but produce a program that does the same thing. I've just bypassed the GPL.

Nope, you've just used that program to create a derivative work, which still falls under the original license.

There's a program that translates code into plain English. You can use this on crypto source, take it out of the US, and translate it back into source. That would bypass the GPL too, since it reduces the code to raw algorithms.

Legally, that too is a derivative work. While it may then be in a form that is acceptable for export, it is still a derivative work.

---

--

DNA just wants to be free...

Some XFS code actually released! by Cef · 1999-08-12 08:58 · Score: 1

At SGI's Open Projects page for XFS you'll find that at LinuxWorld, SGI released documetation, data sheets and some sample parts of the XFS Journalling code (Those parts that are owned by SGI, and which there are no patent issues with).

In other words, SGI is delivering on their Open Source promise with XFS, even if in only a small way (for the moment).

Quoted from their own page... "This initial publication of excerpts of XFS Source code indicates SGI's commitment to port XFS to Linux as GPL code. This code can be freely used with Linux, without the worry of violating the GPL terms by mixing copylefted GPL and copyrighted non-GPL code in Linux. It also shows some interesting code (originally developed for IRIX) which implements the XFS journaling techniques."

Re:What about *BSD? by Luis+Casillas · 1999-08-13 05:58 · Score: 1

The libraries are under LGPL, man.

Readline is under GPL, not LGPL.

Distributing XFS as a plug-in file system separately wouldn't work for M$. You can't use it alone, without the main program. Hence, not "considered independent and separate works in themselves".

I was not talking about the NT kernel, but some BSD kernel. If you took a BSD kernel, and added some compilation option to build it with the GPL XFS code, by using that option you would create a composite work you must distribute under the terms of the GPL. This you can do, since the BSD license allows you to sublicense the source code under the GPL. And because a BSD kernel is something you can use without the XFS code, it is a separate and independent work from the XFS code.

---

This sounds an awful lot like bfs by Anonymous Coward · 1999-08-12 09:17 · Score: 0

hmmm... 64bit journaling file system. sounds quite a bit like bfs (BeOS file system) not that i am complaining or anything, it will be _very_ nice to have a modern fs in both my BeOS and linux partitions. I am wondering though, will it support attributes like bfs, or be framentation proof like bfs. Its a good thing when people arent so proud that they refuse to do something like someone else does, just because someone else did it first... Anyway, this is great! HaveFun! WhiteRabbit

Re:Replace ext2? Yes! Ext2 has 8GB limit + slo fsc by dvdbn · 1999-08-11 23:56 · Score: 1

well, uhh, there is no 8 gig limit on ext2 fs partitions first off... also, on my PII 266 here it does take too much time to fsck my 10 gig /home partition, but it does not take at all near half a day... it takes 4 or 5 minutes.

XFS is coooool tooooo by Anonymous Coward · 1999-08-11 23:57 · Score: 0

I'm not saying it is better, since I havent used JFS, but XFS also has no problem w/ hard power failures.

Re:Why should they cry? They have something better by Passacaglia · 1999-08-11 23:57 · Score: 1

The NT fs arch may or may not be better than vfs, but if no one uses it, it does no one any good. Right now, NT has inadequate choices for file systems, and doesn't seem to be getting any new ones.

JFS .... and LVM? by Anonymous Coward · 1999-08-12 00:01 · Score: 0

Under AIX its the LVM that lets you dynamically resize your filesystems.

I don't believe that the XFS will let you do that under Linux...

Re:What if.... by david@ecsd.com · 1999-08-12 00:05 · Score: 1

Well I imagine that the big tip off would be that Windows suddenly start working right...

Re:Who cares? by Anonymous Coward · 1999-08-12 00:08 · Score: 0

FreeBSD has journal file system. It is broken very badly. In the past of years I have asked when it is going to be fixed. Always I was treated very rudely. I stop using FreeBSD because of these rude peoples. FreeBSD is broke. They do not fix their codes.

D.H. Brown did chew on this by Anonymous Coward · 1999-08-16 04:14 · Score: 0

Back on May 24, 1999. Obscurely buried on their site is a review of SGI's initial decision to open up XFS (PDF viewer required). It does bring up some issues not mentioned in the discussion here so far, like how SGI could make money with the move (mixing a little self-interest with their altruism.)

Re:No it's not, dumbass by howardjp · 1999-08-17 05:24 · Score: 1

The GPL is ugly and immoral. End of story. It also does not promote public domain software. It promotes virus infected software.

Module by Isaac-Lew · 1999-08-12 09:40 · Score: 1

Wouldn't XFS be able to run under *BSD if it were compiled as a module?

Re:Replace ext2? Yes! Ext2 has 8GB limit + slo fsc by Anonymous Coward · 1999-08-12 09:40 · Score: 0

If you think 30 minutes is good, that doesn't mean ext2 is a good thing, it means you have low standards.

Re:JFS is verrrrrrry cooooooool by Anonymous Coward · 1999-08-18 04:12 · Score: 0

Actually WarpServer Advanced gets you HPFS386. For JFS you want the new WarpServer for e-Business (version 4.5/5.0 depending on what docs you read). Up till version 4, the diff between advanced and norm is HPFS386. Now with JFS, you get that at no additional cost so there's just one package. BTW, the only reason for the diff packs before was the $200/server royalty m$ fucked us over with since they hold some right to HPFS386 from way the hell back when they got along.

threading by delmoi · 1999-08-12 10:04 · Score: 1

An OS with threading will be faster on any computer with more than one CPU, with just one CPU it would probably depend on the quality of the programing
"Subtle mind control? Why do all these HTML buttons say 'Submit' ?"

--

ReadThe ReflectionEngine, a cyberpunk style n

Only if ... by Anonymous Coward · 1999-08-12 10:06 · Score: 0

... you're thinking of packaging xfs in a closed-source distribution (a la SunOS, OSF/1, BSD/OS). Lose freedom to ensure freedom or Use freedom to restrict freedom. The world is full of compromises, isn't it ;-)

Getting nitpicky with logic by jwang · 1999-08-12 10:25 · Score: 1

Just for the hell of it,

That would be non sequitur or "does not follow." In essence, you're saying that A leads to B when it really doesn't.

Begging the question is like:

Windows is better than Linux, therefore Windows is better than Linux.

Usually not so blatant but you get the idea.

Jonathan Wang

They admit they can't.... WAS Re:What if.... by Anonymous Coward · 1999-08-12 10:39 · Score: 0

Stallman has admitted that to date, the public pressure has been enough to get people who are in violation to comply with the GPL. Bluff or not, no one has said to the GPL lawyers: "Take us to court." And unless its the universal /. villian of Micro$oft, I doubt we will see any lawsuit. The proof becomes hard any time someone does a re-write of code. Stallman and Co: don't want the 1st case to be a looser.

No it's not, dumbass by delmoi · 1999-08-12 11:08 · Score: 1

If you had half a brain, and had actualy *read* the BSD license you would see that you can redistribute the code or binarys however you want. I could take the sum total of the BSD code and put it under the GPL, or sell a propritary version.
"Subtle mind control? Why do all these HTML buttons say 'Submit' ?"

--

ReadThe ReflectionEngine, a cyberpunk style n

Re:No it's not, dumbass by Brian+Feldman · 1999-08-12 12:48 · Score: 1

No you couldn't. The copyright would have to be reproduced in the license. The GPL is viral, ugly, and completely incompatible with anything other than public domain software, which could be put under the GPL.

--
Brian Fundakowski Feldman

Freebsd is dying by Anonymous Coward · 1999-08-12 11:43 · Score: 1

Why is freebsd dying? Each week brings more bad news to the dwindling number of its fans. Despite the mournful wails of the remaining BSD zealots, the leading industry analysts have coolly concluded that BSD has become a dead end as its user base continues to shrink, and ISVs abandon support. What brought about BSD's headlong spiral of death? What lessons can be learned? Here are some of the reasons put forth for freebsd's failure:

closed development (real contributions limited to old-boy clique).
contempt for others in non-freebsd hacker community.
unrepentant hypocrisy (hatred of the FSF while slurping up all the FSF GNU tools in sight).
rabid jealousy of the success of others (e.g. the success of Linux and BeOS).
fratricidal incompatible fragmentation of BSD community (freebsd, openbsd, netbsd, 386bsd, bsdi).
aging user base (gray beard syndrome).
lack of true innovation (emphasis on trying to match feature check lists).
single source vendor trap.
devoid of third party and ISV support.
death of CSRG and government subsidies.

Freebsd is a classic study in office politics and backroom backstabbing. The freebsd ``core team'' has been devastated in recent months, punctuated by the resignation of many of its senior members. From lax attention to security as highlighted in Usenet's comp.risks, to continued financial woes of a certain vendor of freebsd cdroms, the ongoing decline of freebsd should serve well to instruct others in how not to run a software project.

greedy bastard :) by bfk · 1999-08-12 00:20 · Score: 1

can't you just say thanks for the release of XFS under the GPL, and not some annoying custom license that would make it impossible to fold XFS into the kernel? brian

Re:Who cares? by Anonymous Coward · 1999-08-12 00:22 · Score: 0

Well, at least you were able to make a choice and you'll have to live with it. BSD is fine, but I don't think I'd expect a great deal of 3rd party contribution from the Corps. You see, it is one thing to give stuff to your customers, they call it marketing, advertising, "good will", and just being swell. The non-closing covenents in *PLs are a great way to achive the goal. It is quite another to give it to your competitors, potential competitors, or anyone who may use it against you. In short, we call that a shareholder lawsuit. BSD is the fastest way to get there, while the *PLs effectively forbid it. The GPL, or other *Pls are where the bias will be for Corps that contribute free software. Just as the GPL may be evil for wanna-be grab and make it proprietary types and BSD user camp, the BSD license is poison to most Corp contributors over the long term. BSD's purpose was to allow University work done with public funds to be easily commercialized. The process is simple, professor gets public grant, builds code base, releases the portions built with public money under BSD, and takes the rest private. Its a system origionally designed for the conversion of public funds. It does a fine job at what it does. Corps are NOT universities, they don't think like universities, and they don't operate as such. They have real competitors, present and future, whom they really aren't supposed to aid and abet with dumbfounding willingness. Giving any real or potential competitor a shot at your work product by putting it under BSD is just stupid in the Corps world. So, in making your choices, figure in the fact the a great deal (even most) of the free software work will soon be done by the Corps. And, the *PL licenses are better suited to that reality.

Re:Compatible with SGI xfs partitions by gddavidson · 1999-08-12 00:28 · Score: 1

Does this mean I would be able to just mount the drive from the
200 onto a Linux (i386) box and pretty much be all set?

I think this is exactly what it means. We are also looking forward to being able to do this. I'd love to buy a lower cost Linux box from SGI (although I can't say enough good things about the Origin.) Or put one together myself.

Re:Who cares? by Ethan · 1999-08-12 00:30 · Score: 1

(a) Who cares?

I do, and you should. Just because you don't use *BSD, and I don't use *BSD (although I have OpenBSD installed, and I like what I've seen, I use Linux as my primary OS) doesn't mean that we should not be concerned with the welfare of the *BSD movement. You do not appear to be aware of this, but Linux has gotten a LOT of good out of the BSD developers and BSD code over the years. I know that for a long time (and I believe it still does) the Linux kernel booted and displayed a message to the effect of "This program contains code Copyright (c) the Regents of the University of California, Berkeley ..." You might recognize "Berkeley" as the B in BSD and that snippet (which may not be 100% word for word) as part of the BSD license. Many of the utilities you use every day on your Linux system are *BSD utilities, or derived thereof.

Be careful who you scorn, especially if you haven't done your research. Remember that competition is a good thing, and that the *BSD people are pursuing a goal, while not identical, similar to that of the Linux crowd.

(b) You can't integrate other Linux kernel enhancments into BSD

While this is true (although there may be ways to work around it ?), that does not mean that we should not work for a solution equitable to all parties.

(c) SGI is getting into Linux. It doesn't give a damn about BSD (as most of us don't)

I can't speak for SGI, and while I cannot technically speak for "most of us" (as I am sure you cannot), I do not believe this is true. There are many people on Slashdot who care as little for Linux as you seem to care for *BSD. There are also many who feel as I do, using Linux day-to-day but following the *BSD activity and applauding those developers for their efforts, or vice versa.

(d) Freer license would mean it could get into NT as well

And what is the problem with that? You (as well as many others) seem to miss the point that a stable and functioning NT would be nothing but a boon to the industry. You rant against Microsoft about the instability and uselessness of their operating system, and then seek to deny them the tools to fix it.

(e) If BSD really wants it, BSD can GPL itself

While I don't know the particulars of the BSD license enough to comment on the truth of that statement, I can comment on its invalidity for other reasons. *BSD is under the license that it is under because the authors believed that that license was the one it should be under. Just like you belive the GPL to be superior, they believe the BSD license to be superior. I can see the merits of both. Why don't you release your software under the *BSD license instead of the GPL (if you have any, which I doubt)? That is the same as what you would ask them to do...

I would like to think that you are just misguided and not a troll. If you are misguided, think before you speak. If you are a troll, go home.

Ethan

Re:Who cares? The Corps do. by Anonymous Coward · 1999-08-12 00:31 · Score: 0

Well, at least you were able to make a choice and you'll have to live with it. BSD is fine, but I don't think I'd expect a great deal of 3rd party contribution from the Corps.

You see, it is one thing to give stuff to your customers, they call it marketing, advertising, "good will", and just being swell. The non-closing covenents in *PLs are a great way to achive the goal.

It is quite another to give it to your competitors, potential competitors, or anyone who may use it against you. In short, we call that a shareholder lawsuit. BSD is the fastest way to get there, while the *PLs effectively forbid it.

The GPL, or other *Pls are where the bias will be for Corps that contribute free software. Just as the GPL may be evil for wanna-be grab and make it proprietary types and BSD user camp, the BSD license is poison to most Corp contributors over the long term.

BSD's purpose was to allow University work done with public funds to be easily commercialized. The process is simple, professor gets public grant, builds code base, releases the portions built with public money under BSD, and takes the rest private. Its a system origionally designed for the conversion of public funds. It does a fine job at what it does.

Corps are NOT universities, they don't think like universities, and they don't operate as such. They have real competitors, present and future, whom they really aren't supposed to aid and abet with dumbfounding willingness. Giving any real or potential competitor a shot at your work product by putting it under BSD is just stupid in the Corps world.

So, in making your choices, figure in the fact the a great deal (even most) of the free software work will soon be done by the Corps. And, the *PL licenses are better suited to that reality.

Re:The real reason... by Agent+Drek · 1999-08-12 01:58 · Score: 1

uhh as far as I can tell your friend was running
/usr/freeware/bin/ls --help

which says at the bottom:

"Report bugs to fileutils-bugs@gnu.ai.mit.edu"

instead of
/sbin/ls --help

which yields
[drek@nessie] ~ > ls --help
Illegal option -- e
Usage: ls -RadC1xmnloghrtucpFbqisfLAHMDPSj [files]
Exit 2

later,
-derek

Re:Who cares? by platypus · 1999-08-12 01:58 · Score: 1

Ethan,
I agree with most of what you say. But theres one
problem in todays OS-industry. From my expirience
the main point to get people to use linux or bsd
is the superior stability. If a company (nowadays
its MS, but it could have been i.e. ibm) has a
defacto monopoly and a very strong public image,
the only way to compete is in fact to be cheaper AND better.
If one uses bsd-license, he gives away
all his advantages but gains nothing when it comes
to easily portable features like filesystems etc..
So, while I wish the BSD-people the very best, I
believe in fact that in such wierd circumstances
as todays IT-industrie, GPL is the way to go for
many things.

Re:Who cares? by Anonymous Coward · 1999-08-12 00:38 · Score: 0

thats a totally *stupid* statement. youre either blinded by the fact youre a BSD user or youre an idiot. The point of the GPL is to prevent commercial companies from abusing the code - and thats a damn good point. GPL code is truly open source - and no amount of whining from lame BSDers is gong to change that.

Untouted feature could make XFS ideal for Darwin by Valdrax · 1999-08-12 00:39 · Score: 1

I was reading the white paper on XFS when I noticed a section that had some information which might be useful for Mac OS X:

3.2.5 Attribute Management

XFS supports user defined Arbitrary Attributes which allow information about a file to be stored outside of the file. For example, a graphic image could have an associated description or a text document could have an attribute showing the language in which the document is written. An unlimited number of attributes can be associated with a file or directory. They can have any name, type or size of value. These Attributes represent the first substantial enhancement to the UNIX file interface in 10 years.

Wouldn't this be a great way to preserve the Mac OS idea of resource forks? Any ideas?

--
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").

Re:Request for info on ext3 by rbf · 1999-08-12 02:15 · Score: 1

I would also like some in depth info on ext3, and what would be nice is a comparison of ext2, ext3, and SGI's XFS. Thanks! rbf ALPHA LINUX POWERED!

Re:The Viral Nature of GNU (Not Really) by Shabazz · 1999-08-12 00:41 · Score: 1

This is wishful thinking. If someone sued MS for using GPL'd code without releasing their source the court would not (could not) force MS to release their source code.

The remedy would most likely be a injunction from selling software in violation of the license and money award for damages already done. Just because the license says the code must be released doesn't mean the ct will do so.

For instance, if the GPL said that violators will be shot and killed, do you think the court would enforce it? Probably not.

Re:What about *BSD? by Luis+Casillas · 1999-08-12 02:19 · Score: 1

I'll ignore the "truely [sic] free" part and just concentrate on the license issues.

I'll quote the GPL, section 2:

If identifiable sections of that [composite] work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when youdistribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it.

What this means is the following:

If someone were to integrate the BSD kernel with GPL XFS code, and distribute this as a single package, that particular BSD kernel package would have to be distributed in compliance both with the terms of the GPL and BSD licenses at once. In practice this is doable (and the FSF has done it, BTW); it's just a GPL + a BSD advertising clause.
If someone were to create an add-on package that allowed people to take the BSD kernel package + a XFS package for it, distributed under the GPL, and compile BSD kernels with XFS support, this could not affect the distribution terms of the GPL code. Stuff like this is done all the time -- look at all the non-GPL free programs that allow you to optionally link them to GPL libraries like Readline (Python and the Haskell interpreter Hugs come to mind).

Thus, GPL XFS code could be integrated into BSD without tainting its license, since the BSD code is a separate work. What you can't do is distribute the combination under terms incompatible with the GPL.

---

Re:The real reason... by ChrisKnight · 1999-08-12 02:21 · Score: 1

:SGI's //bin/ls does NOT accept a --help flag, nor does it pop up a GPL notice.

I DID NOT say /bin/ls! Although I was wrong anyways, I said it was the custom XFS ls that ships with XFS, NOT /bin/ls. (And I was wrong, sorry.)

It is actually the custom ls that ships with the DMF product. If you execute 'dmls --help' you get a help screen with a GNU license statement.

My apologies for confusing this with the XFS release to GPL. I'm not an SGI or IRIX user and I botched this up. Sorry.

But if you are running DMF, give it a try. I understand that SGI is fixing this 'mistake' in the next release.
-ck

--
-- This sig is only a test. If this were a real sig it would say something witty. --

Re:The real reason... by ChrisKnight · 1999-08-12 02:22 · Score: 1

:SGI's //bin/ls does NOT accept a --help flag, nor does it pop up a GPL notice.

I DID NOT say /bin/ls! Although I was wrong anyways, I said it was the custom XFS ls that ships with XFS, NOT /bin/ls. (And I was wrong, sorry.)

It is actually the custom ls that ships with the DMF product. If you execute 'dmls --help' you get a help screen with a GNU license statement.

My apologies for confusing this with the XFS release to GPL. I'm not an SGI or IRIX user and I botched this up. Sorry.

But if you are running DMF, give it a try....

-ck

--
-- This sig is only a test. If this were a real sig it would say something witty. --

Re:A pre-emptive strike [OFF TOPIC] by Hackboy · 1999-08-12 02:30 · Score: 1

"But always she's the spectre of uncertainty I first endured, then faded, then embraced..."

Wow, I didn't realize there were any other people on Earth that listen to Toy Matinee. One of my favorite albums of all time.

Would be nice by Scudsucker · 1999-08-12 02:41 · Score: 1

But I don't think that Apple would switch to XFS this late in the development cycle of OSX; it would be a major change in how the OS handles files and addopting XFS as default would delay X's release.

OTOH, maybe Apple got some inside info from SGI last year when Apple started to think about licensing OpenGL and will adopt XFS as default.

Re:Violators will be shot and killed by bcaulf · 1999-08-12 12:01 · Score: 1

Your suspicions are all the more reason why you should join the fsf cadres. Code can only free so much, you need small arms to put the teeth behind the GPL.

Whoever dies with the most filesystems wins. by acb · 1999-08-12 12:09 · Score: 1

The more the merrier. If you have several dozen filesystems to choose from, each with its own unique features (some historical, some futuristic), that is a good thing.

Of course, most people will probably only use one or two general-purpose filesystems, though the others will find their niches.

Look at the site by acb · 1999-08-12 12:33 · Score: 1

If you looked around oss.sgi.com, you'd notice that there are lots of OSS projects. For example, there have been about 20 patches to the Linux kernel, mostly dealing with performance enhancements; some of these are already in the 2.2 and 2.3 trees.

Re:The real reason... by Brian+Feldman · 1999-08-12 12:46 · Score: 1

What this really tells me is that IRIX comes with a buggy getopt().

--
Brian Fundakowski Feldman

bsd faces new lawsuit by Anonymous Coward · 1999-08-12 13:30 · Score: 0

not directly related to this discussion; new lawsuit is going to be opened against bsd and berkeley for patent violation. bloomberg webcast mentioned this earlier today. some patent rights now belong to startup utah company. plaintiff seeks immediate injunction against bsd distributers. bad bad news.

Jorunallings FS for Linux already on the cards? by RallyDriver · 1999-08-12 14:15 · Score: 1

I thought someone in RH's employ (at risk of naming names, someone based in Edinburgh) was writing a journalling add-on for ext2/ext3 ?

Re: No for JFS too by Anonymous Coward · 1999-08-12 14:17 · Score: 0

JFS is not Veritas filesystem. Sun uses Veritas to be able to expand filesystems on-the-fly. This is default in AIX. I think the concept of "journaling" was first introduced in JFS (hence it's name). In another response it's says XFS is not licensed from Veritas, so your post looks very wrong, and if you think about it. How could SGI make a filesystem GPL which they have bought from another company ?

Re:Who cares? by howardjp · 1999-08-12 01:04 · Score: 1

FreeBSD does not have a journaling FS. FreeBSD does support FFS, MFS, NFS, FAT, VFAT, NTFS, Coda, and ISO9660. None of these are jounraling.

Re:Replace ext2? Yes! Ext2 has 8GB limit + slo fsc by Robin+Hood · 1999-08-12 01:07 · Score: 1

As someone else already pointed out, there's no 8GB limit on the size of an ext2 partition, and the fsck isn't all that slow. I've got a 54GB RAID partition (split across three drives) here (not my home computer, though. I wish! :-) ), and fsck'ing it takes about half an hour (I timed it once).

I still am looking forward to putting xfs on that partition, though.
-----

--
The real meaning of the GNU GPL:
"The Source will be with you... Always."

Reply to the flamebait by ashpool7 · 1999-08-12 01:11 · Score: 1

My programming skills have yet to enter into the experience required to write a new filesystem. However, when I do get to that level, and such a filesystem is not implemented, I will write such a thing.

I am not a lazy ass, just not educated enough. I was wondering if anyone had plans, or would like to make them, to bring such a filesystem to light in a timeframe earlier than is possible for me.

Re:Who cares? by Ded+Bob · 1999-08-12 02:49 · Score: 1

OTOH, Microsoft has a million "monkeys". If they can look at the source, they can basically just duplicate all the ideas found in it with the man-power they have. To Microsoft, just being able to see the source would be equivalent to releasing it under the BSD license.

Re:Who cares? by howardjp · 1999-08-12 01:13 · Score: 1

You've been blinded by brainwashing of the FSF and Richard Stallman. Anyone who has read the GPL can tell you it is even more encumbered than any commercial license. At least with a commercial product, I can purchase rights to redistribute the program without source.

The Berkeley and X licenses are the only truely free licenses in common use.

Re: O200's by Cookie+Monster · 1999-08-12 02:49 · Score: 1

We run a few 200's as well. Would definatly be
nice to be able to swap the drives willy-nilly.
Still havn't figured out if the 200 fully supports hotswaping of the hard drives. Linux is not ready
for that yet either anyway.
XLV would be more usefull still. Striping is the
only thing that keeps the sgi machines usefull -
the 180-250mhz processors don't go thru cgi
as fast as the new pentiums. (IMHO)

The real reason... by ChrisKnight · 1999-08-12 01:17 · Score: 1

SGI had NO choice. They HAD to GPL XFS.

A friend of mine at NASA was playing with the 'custom' ls that ships with XFS, and since the docs suck he fed it '--help' as an option. Up popped a help screen topped with a GPL license statement.

XFS was using 'stolen' GPL code, so they either had to completely rewrite it or give it away.

They hide that little tidbit, don't they?

If you are running XFS, try tagging the above and see what happens. (Sorry I don't remember the command itself, I don't get to play with IRIX boxes much.)

-ck

--
-- This sig is only a test. If this were a real sig it would say something witty. --

Re:The real reason... by ChrisKnight · 1999-08-12 03:39 · Score: 1

: uhh as far as I can tell your friend was running
: /usr/freeware/bin/ls --help

Actually, he was running dmls --help

Ya know, I'll never post to slashdot again before I have had my second cup of coffee. Making a mistake here is worse than getting into a fender bender with an undercover cop. :)

-ck

--
-- This sig is only a test. If this were a real sig it would say something witty. --
Re:The real reason... by cweber · 1999-08-12 01:54 · Score: 1

That is of course complete bullshit. Irix 6.5 includes a lot of GNU tools in /usr/gnu. Don't remember whether it's part of the default install or not. GNU tools are also distributed on SGI's Freeware CD and installed into /usr/freeware. Your friend most likely had /usr/gnu or /usr/freeware in his path before /bin and /usr/bin, as many of us do to get the benefit of better tools. He or you should check facts and configs before spouting such nonsense. SGI's //bin/ls does NOT accept a --help flag, nor does it pop up a GPL notice.
Re:The real reason... by Anonymous Coward · 1999-08-12 15:51 · Score: 0

Be freaking funny if they covered up all the includes, and somehow completely missed hiding a help screen though... :-)

Re:What about *BSD? by Anonymous Coward · 1999-08-12 02:50 · Score: 0

Somebody can port it as an optional loadable kernel module for BSD. The only thing they will have to watch out for is to avoid adding code under a BSD advertising clause to the XFS code, since then the combined code could not be distributed under GPL; this shouldn't be a problem if they can stick to mostly using headers, etc. Then they can distribute a GPLed version for their favorite BSD kernel. (though the major BSD projects probably wouldn't accept it as a core component :)

Re: Ext2 has a 4 TB limit. by Anonymous Coward · 1999-08-12 03:04 · Score: 0

And files arent bigger than 2 gigs. xfs has a 8 EB limit. (A Exa is a million Tera) Frank

ACLs? by Anonymous Coward · 1999-08-12 03:05 · Score: 0

Does anyone know whether XFS supports access control lists? A lot of people want a filesystem that supports more specific permissions than User/Group/Other Read/Write/Execute. There could be more specific permissions, such as:
- read
- modify (without making the file larger)
- append (can't modify existing information)
[modify+append would be the same as ext2 write permission]
- execute
- delete (for files or directories)
- create (directories only)
- modify ACL

I know NTFS supports some of these things, but it would be nice to have something similar on Linux. The "attribute management" feature of XFS could probably be used to hold an ACL.

Re:What if.... by zmooc · 1999-08-12 03:18 · Score: 1

How do you prove that Win2K doesn't use some modified Linux IP stack?
Well I don't know much about it, but remember Teardrop?:) That DOS-attack took down nearly all OS'es. So they all had the same bug. Coincidence? Think not.

--
0x or or snor perron?!

um no.. by Anonymous Coward · 1999-08-12 03:19 · Score: 0

the world might be a better place but it would be a blow to free software. people will think: hey, im making this stuff for free so someone else can sell it? wtf! I'll sell it myself. fsck them and free software! this will turn the free software movement sour.

Re:um no.. by Anonymous Coward · 1999-08-12 15:46 · Score: 0

BSD seems to be doing okay.

I'm sorry, that is just wrong. by MenTaLguY · 1999-08-12 03:21 · Score: 1

Anyone who has read the GPL can tell you it is even more encumbered than any commercial license. At least with a commercial product, I can purchase rights to redistribute the program without source.

This is possible with GPLed software as well. The GPL does not (and indeed cannot) tie the hands of the copyright holder(s) with respect to licensing.

Just like any other software, it is entirely possible for you (as an individual or as a representative of a corporate entity) to negotiate alternate licensing arrangements with the copyright holders, which may or may not involve a transfer of funds. This is true regardless of the license attached to specific releases of the code.

The Berkeley and X licenses are in essence no different in this regard; they simply grant that permission to relicense implicitly.

---

--

DNA just wants to be free...

Re:I'm sorry, that is just wrong. by howardjp · 1999-08-12 03:48 · Score: 1

I cannot purchase the rights to dialog. The original author is unknown. But it has a GPL at the top. Looks like I just need to rewrite it myself.

This problem also extends to large GPL'd products. If you own the licensed on xxx.c and xxx.h and roblimo owns abc123.c, def456.c, and cmdrtaco owns 987654321.c, and you won't sell to me, I'm
screwed.

These problems are handled by X and BSD licenses and no duplication of effort is required.
Re:I'm sorry, that is just wrong. by Anonymous Coward · 1999-08-12 04:51 · Score: 0

NO. simply release your code under the GPL. Then you can redistribute both source and binary freely under the terms of the GPL. I recommend you actually read the GPL before spouting crap. Note that the GPL doesnt prevent you from selling your product, just from violating the original authors intent to release the code in a proprietary manner. If you want to release in a proprietary fashion - you gotta rewrite it.
Re:I'm sorry, that is just wrong. by MenTaLguY · 1999-08-12 05:11 · Score: 1

I cannot purchase the rights to dialog. The original author is unknown. But it has a GPL at the top. Looks like I just need to rewrite it myself.

Hrm; that's an interesting situation. Is the license even valid when there is no copyright information? (probably, but it's worth looking into). This is a valid point, except it is not a problem specific to the GPL.

This problem also extends to large GPL'd products. If you own the licensed on xxx.c and xxx.h and roblimo owns abc123.c, def456.c, and cmdrtaco owns 987654321.c, and you won't sell to me, I'm screwed.

You get the exact same situation when you deal with software under other licenses with a number of copyright holders. You can just purchase license to roblimo's and cmdrtaco's code, and reimplement xxx.h and xxx.c yourself. That's still considerably less work than rewriting from scratch.

This, too, is not a problem specific to the GPL. For instance, Netscape had to rip out large chunks of Mozilla before releasing the source, because not all the different copyright holders would agree to the relicensing.

These problems are handled by X and BSD licenses...

The X and BSD licenses take the expediency of explicitly granting open-ended permission to relicense, avoiding the necessity of negotiating with the copyright holder. That's not a bad thing in and of itself, unless you like your code being enhanced and used in a proprietary fashion without any compensation to you or the other users/contributors, who had been giving to each other freely.

There's a difference between giving freely and and letting yourself be taken advantage of.

...and no duplication of effort is required.

This is assuming that everyone who uses the BSD/X-licensed code relicenses it in a compatible manner. That is not always the case.

To provide a counter-example to your assertion, let's say entity X takes your software package Y, and releases an enhanced version under license Z (which you cannot personally accept). If you want the new features under an acceptable license, you'll still have to reimplement them yourself. It's even harder if license Z means a completely propreitary licensing arrangement, such that you have to reverse-engineer the features they added. Thus, BSD/X licenses may (indirectly) result in duplication of effort, too.

Every license has its own set of problems and potential pitfalls.

---

--

DNA just wants to be free...

Re:Untouted feature could make XFS ideal for Darwi by howardjp · 1999-08-12 03:24 · Score: 1

It would probably be an excellent way to handle resource forks, however, the GPL will prevent that from ever happening.

Re:geez, enough complaining by rew · 1999-08-12 03:36 · Score: 1

> and now they're freeing a huge, important
> chunk of code that is desperately needed
> and would be a massive development effort
> to start from scratch!

At 7000 lines of code, it may be huge, but it certainly isn't a "huge part". It represents about 7% of the code. There is a total of about 100k lines. 13 times more to come....

It isn't the easiest code to read. I like code with funny comments. This looks as if the PHB types got to read the code regularly. Humor is important. Not much of that in this code.

Oh well.....

-- Roger.

(computer needs vacuuming again: my X server just crashed ;-( )

What if.... by gsaraber · 1999-08-11 22:27 · Score: 2

So SGI will add XFS to the Linux Kernel shortly, great, but what if some company steals stuff from the XFS technology by looking at the kernel source and integrates it into their product?
I know that the GPL is supposed to prevent that, but the thing is, how do you prove it?
example: how do you prove that Win2K doesn't use some modified Linux IP stack? Nobody's allowed to see the source so nobody will ever find out right?

guess i'm just paranoid :-)

Re:What if.... by mentaldent · 1999-08-12 15:39 · Score: 1

>example: how do you prove that Win2K doesn't use some modified Linux IP stack? Nobody's allowed to
>see the source so nobody will ever find out right?

Someone's already thought of that.

An excerpt from http://www.phrack.com/search. phtml?view&article=p54-9 which describes nmap, an OS fingerprinting-by-TCP/IP-stack-details tool:

TCP Initial Window -- This simply involves checking the window size on
returned packets. Older scanners simply used a non-zero window on
a RST packet to mean "BSD 4.4 derived". Newer scanners such as
queso and nmap keep track of the exact window since it is actually
pretty constant by OS type. This test actually gives us a lot of
information, since some operating systems can be uniquely
identified by the window alone (for example, AIX is the only OS I
have seen which uses 0x3F25). In their "completely rewritten"
TCP stack for NT5, Microsoft uses 0x402E. Interestingly, that is
exactly the number used by OpenBSD and FreeBSD.

-----------------

Interesting indeed! Hmmm, looks like MS has been caught with their pants down and their finger in the pie and their hand in the till. :)

mentaldent
Re:What if.... by Anonymous Coward · 1999-08-12 15:47 · Score: 0

Um, that wouldn't be really hard to change. You could change a single case or two. I mean, it's not like you'd have to recode much.
Re:What if.... by superlame · 1999-08-12 01:44 · Score: 1

The GPL is not designed to prevent code from being borrowed. It is designed to allow it. So, there is not a thing keeping the FreeBSD developers from adding XFS support to their kernel, or to keep them from copying the linux kernel code that allows Linux to run on SGI intel machines.

In the case a Microsoft and the IP stack, they wouldn't steal the linux IP stack, they would take the FreeBSD stack. And even if they didn't release the source, it would be legal, because that is the way the BSD license works.

If they were to steal the linux IP stack, then if there is any reasonable reason to suspect they did so, the the MS code can be supeaned (sp).

--
-- Superlame http://catpro.dragonfire.net/joshua/
Re:What if.... by Anonymous Coward · 1999-08-12 03:48 · Score: 0

The GPL is not designed to prevent code from being borrowed. It is designed to allow it. So, there is not a thing keeping the FreeBSD developers from adding XFS support to their kernel, or to keep them from copying the linux kernel code that allows Linux to run on SGI intel machines.

Your wrong, the BSD people cannot use GPL'ed code without changing their licence to the GPL.

If I made a licence called the NGPL (new GPL) that was word for word identical to the gpl, I still would not be able to use GPL code in it.
Re:What if.... by Anonymous Coward · 1999-08-12 15:55 · Score: 0

Yeah, I read about that. Most TCP stacks around these days are based on some old UNIX stack...I dunno, guess it was Berkely or something. Routers, too.
Re:What if.... by Anonymous Coward · 1999-08-12 15:56 · Score: 0

Ya, so? So they based it on BSD. Big deal. And the GUI was ripped off and tossed around. And everything else. BSD didn't say they couldn't.
Re:What if.... by Anonymous Coward · 1999-08-12 05:17 · Score: 0

Not likely. Large corporations have legions of lawyers paid to be paranoid about such things. It's hard enough to get =legal= uses of GPL software out, much less illegal ones. If anybody is misusing GPL software it is likely a small company that is more desperate than paranoid.
Re:What if.... by CaseyB · 1999-08-11 23:08 · Score: 1

So SGI will add XFS to the Linux Kernel shortly, great, but what if some company steals stuff from the XFS technology by looking at the kernel source and integrates it into their product?

Then the world is a better place.

The free software movement was founded on the notion that sharing ideas and information -- amongst everyone is a Good Thing. The GPL is not designed to prevent people from learning something from the code. It's designed to do the opposite.

*heavy sigh* by Anonymous Coward · 1999-08-12 15:15 · Score: 0

Since when have "geeks" become obsessed with licensing issues and simply overlooked the new features available. SGI spent a *lot* of money developing XFS. I have used it for quite some time and have *never* lost data, and even after a power outage had a box with 18 gig of disk come back up in 1 minute. Now, they're turning around giving this away for free, and reading these comments, I see very little laud, a bunch of complaining about the very existence of XFS as a choice for Linux users (!!!), and then a bunch of pedantic griping about licensing issues? I always thought the canonical geek was a lot more into the software. You're all starting to sound like paranoid, delusional MIS mavens.

What about *BSD? by Anonymous Coward · 1999-08-11 22:31 · Score: 0

Won't Releasing this under GPL cause problems with intergrating with truely free OS's such *BSD's?

Re:What about *BSD? by drwiii · 1999-08-12 01:25 · Score: 2

If BSD wants XFS, they'll port it. It'll just reside as an option in the non-free section of the installation.
Re:What about *BSD? by Anonymous Coward · 1999-08-11 22:52 · Score: 0

away troll! away, i say.
Re:What about *BSD? by nikc · 1999-08-11 22:52 · Score: 1

AC wrote:

Won't Releasing this under GPL cause problems with intergrating with truely free OS's such *BSD's?

Possibly. There's a thread about this just started on the freebsd-hackers mailing list.

Matthew Alton is the person to talk to -- I'll not include his e-mail address here for spam reasons, but the post in which he announces he's working on it is http: //docs.freebsd.org/cgi/getmsg.cgi?fetch=269893+0+c urrent/freebsd-hackers.

You can follow the thread from that link.

N
Re:What about *BSD? by X · 1999-08-11 22:54 · Score: 1

Yes, and this is exactly why SGI released it under GPL, and exactly why GPL'd software is going to have better luck getting corporate support.

Think about it for a moment... if XFS is released under a BSD license, how long before all of SGI's competitors (including MS's Windows NT) have ported the code and added it to their own OS's? At least with GPL'd software, if they're going to do it, they have to publish their own kernel as well.

--
sigs are a waste of space

Replace ext2? Yes! Ext2 has 8GB limit + slo fsck. by Anonymous Coward · 1999-08-11 22:32 · Score: 0

With 20GB hard drives falling below the $250 mark, xfs will be able to do partitions of over 8GB without forcing Linus users to resort to such gawdaful things as vfat when large partitions are called for. This is the minor benefit of xfs, though.

xfs recovers faster from and is far less prone to getting screwed up from SUDDEN POWER OUTAGES for those of us not running UPSs on out Linux boxes. Ever run fsck on an 8GB filesystem? There goes half your day. xfs will come back up a lot faster and a lot more intact.

I'm not waiting for the distros to bundle xfs. I'll be compiling it in the second it becomes available.

What are you smoking? by Anonymous Coward · 1999-08-11 22:33 · Score: 0

Whatever it is, can I have some? You're comparing XFS to reiserfs and questioning whether it's a good idea or not to have XFS. You either don't know what XFS is, or you don't understand the characteristics that distinguish it from reiserfs.

Re:Replace ext2? Yes! Ext2 has 8GB limit + slo fsc by Anonymous Coward · 1999-08-12 15:37 · Score: 0

Everything is relative. ext2 beats the snot outta FAT. That makes me happy. If RH comes along with RH Linux 7.0 that includes xfs or whatever and it's *even better*, I'll be even happier. :-)

Microsoft, read this and weep. by umoto · 1999-08-11 22:35 · Score: 1

See how an open-sourced base platform can adapt to changing needs? Even such a fundamental part of the operating system can quickly be replaced with something better. Go SGI & Linux!

The real question is; HOW MANY DEVELOPERS! by law · 1999-08-11 22:44 · Score: 2

How many developers will SGI bring to the table? One? or thirty?
That will realy show me, SGI has less resorces then ever before.
How much are they going to invest in the future? they have allready paid for XFS.
What will they do? just maintain the code and stand back? Or will the actively develop on all of the Linux Kernel? or just part?
That will truly show the heart of SGI.

--
"Think of it as evolution in action."

Avoid NIH syndrome by squarooticus · 1999-08-11 22:45 · Score: 1

Now, if we can all avoid the NIH ("Not Invented Here") syndrome, it might actually make it into the main kernel distribution.

Kyle

NP: Queensryche, Empire
--
Kyle R. Rose, MIT LCS

--
[ home ]

Re:We take em to court by Anonymous Coward · 1999-08-12 15:45 · Score: 0

No, st00pid person. He was using it as an example.

BSD kicks GPL butt by Anonymous Coward · 1999-08-12 15:45 · Score: 0

You're thinking of BSD. Which actually *does* promote sharing of information.

Request for info on ext3 by John+Kacur · 1999-08-12 01:38 · Score: 1

I noticed in the comments on the Linux Today site, people were discussing ext3 development. Can anybody point out where to find more information about his, ie web sites or usenet groups?

I don't see a problem. by Anonymous Coward · 1999-08-12 03:41 · Score: 0

If the FreeBSD community wants to use GPL'd code, then they should switch to the GPL license. I'm not sure I understand what the problem is here.

Re:Who cares? by Anonymous Coward · 1999-08-12 03:53 · Score: 0

Yup. The GPL is designed to not allow anyone to use it's code without that broject being GPL'ed.

Re:Who cares? by platypus · 1999-08-12 04:11 · Score: 1

Well, they could also just hire the developers (which they have before) to get the know-how, or buy a company which has analogous know-how.
But the important thing is marketing, look at the comments in all the treads about xfs. How many people here are really in need of a journalling filesystem? How many people do even know what journalling is? There is this babbling about this cool sgi thing, which will now be available for linux too, yeah.
Anyway, with or without source, Microsoft could get something which is _like_ xfs, the point is that they are not able to say: "Hey, we have xfs too, based on the original sources.".
They will have at least to pretend they reengineered(sp?) it, and everyone will think this feature has the burden of being newly developed (esp. by Microsoft) from scratch. The shiny feature of using an established product is disappeared.

Color-ls under IRIX? by Rational · 1999-08-12 04:20 · Score: 1

Speaking of ls and SGI... Does anybody know of a ls that supports color (like under Linux) and compiles on a SGI box?

Thanks...

--
"Be nice, veer left, and never stop thinking" Iain Banks - Walking On Glass

JFS is verrrrrrry cooooooool by Anonymous Coward · 1999-08-11 22:47 · Score: 0

Once apon a time, in a galaxy far, far away IBM had an x86 version of AIX. Only ran on a few certain oddball MCA-bus PS/2s however. Yechhhhh! Dunno if that early ver of AIX had the JFS that we've all come to know and love tho', but it would be cool if they'd port 4.3 to the x86 platform so we could run it on Athlons :-). I really love to be able to dynamically grow filesystems, lv's and vg's on the fly while the system is up and in use. I really hate the RS/6K hardware tho', it's mondo overpriced for what it is and leave it up to IBM to proprietarize a standard wide-SCSI connector such that only an IBM brand SCSI cable will fit it. :-( Those kinds of things really irk me.

Compatible with SGI xfs partitions by Anonymous Coward · 1999-08-11 22:48 · Score: 0

We run webservers on Origin 200's, using external disks in xfs format. We're starting to use Linux as our webserver more and more now. Does this mean I would be able to just mount the drive from the 200 onto a Linux (i386) box and pretty much be all set?

Plenty by Anonymous Coward · 1999-08-11 22:56 · Score: 0

SGI is getting into this big time. http://pmitros.mit.edu/sgistrat

Why should they cry? They have something better. by Anonymous Coward · 1999-08-11 22:57 · Score: 0

Sorry to rain on your parade, but the NT file system architecture kicks ass compared to linux or BSD VFS-based architectures. MS can plug in any filesystem they care to write relatively easily. Of course, so can any modern UNIX. Unfortunately, NT is *WAY AHEAD* when it comes to the interface presented for file system authors. The only advantage that BSD or linux has here is cost; in order to write an NT filesystem, you need to shell out US$1000 just for the SDK.

A bit of info about file journaling by ChrisRijk · 1999-08-11 22:58 · Score: 2

This is a LONG article from SunWorld mag, and you have to go down a fair bit to get to the info about file journaling, and such, but it's a good read:
Getting to know the Solaris filesystem, Part 1

Tweet, troll. by bkosse · 1999-08-11 22:59 · Score: 1

intergrating with truely free OS's

Knock it off.

--

--
Ben Kosse
Remember Ed Curry!

Guaranteed I/O rate on XFS by Anonymous Coward · 1999-08-12 04:23 · Score: 0

I'd expect that's dependent on hardware support. Meaning don't expect to see it on your average PC linux box.

Re:Violators will be shot and killed by Anonymous Coward · 1999-08-12 15:59 · Score: 0

The GPL Shoot on Sight squad wants *you*!

Roving teams of heavily-armed Linux fanatics cruise the neighborhood looking for GPL violators. The goverment is paralyzed (Flash to car of armed people with bumper sticker "Try enforcing crypto bans NOW!")... :-)

ex2fs to xfs conversion? by Anonymous Coward · 1999-08-12 04:30 · Score: 0

Great! can't wait to get it, I have had some realy bad experiences with non-Journaling filesystems.

But any chance of a ex2fs to xfs conversion utility, so we don't have to backup;mkxfs;restore?

I have a bunch of boxes running Linux, and it would take a lot of hassle to convert them without a conversion utility.

I have never used an IRIX before, but IBM's JFS implementation on AIX and OS/2 is great!

Re:We take em to court by Anonymous Coward · 1999-08-12 04:36 · Score: 0

Are you people retards or something? Anyone with a brain in their head knows that M$ got their original TCP stack from BSD. (They even called it Berkeley TCP/IP for a while)

This is the end of BSD by Anonymous Coward · 1999-08-12 16:02 · Score: 0

Well, had to come around sooner or later. Sorry to see you go...but not too sorry. Ha! (Always laughing about a jerky Linux cursor, huh...that'll teach 'em)

XFS/JFS/FSCK by Anonymous Coward · 1999-08-12 04:48 · Score: 0

When? Advantages (other than no wait for fsck, which can be hours)?

Anti-MS moron --NTFS *is* journaling by Anonymous Coward · 1999-08-12 16:08 · Score: 0

NTFS supports journaling just fine. A quick Altavista search:

-"The Windows NT File System (NTFS) is an advanced file system that supports journaling, extremely large storage media, and long file names. NTFS provides the highest level of security and file recovery, and is the file system of choice on Windows NT."

Duh. Just because it's Microsoft doesn't mean that it doesn't beat the snot outa the Linux/*BSD equivalent. Sorry, Linux/*BSD fans.

Gee, and Win2k has a better TCP stack, too...hmm, I wouldn't be betting on *NIX any time soon...

Re:Who cares? by Anonymous Coward · 1999-08-12 16:10 · Score: 0

Restrictive open source, but open source.

That FSF stuff is a bunch of propoganda. "Free speech" my foot. *There are more restrictions in the GPL than the BSD licence.* End of story.

I can prove that BSD is doomed by Anonymous Coward · 1999-08-12 16:13 · Score: 0

BSD can -> GPL
GPL cannot -> BSD

A one way valve. Eventually everything will be GPL.

Re:Who cares? by Ded+Bob · 1999-08-12 05:03 · Score: 1

Microsoft could get something which is _like_ xfs, the point is that they are not able to say: "Hey, we have xfs too, based on the original sources.".

I believe Hell will freeze over before Microsoft brags about using anyone else's source. Microsoft invented sockets, remember? Winsock instead of Berkeley sockets. If they even used the source code to XFS, they would rename it the WinFS or something like that.

I still think that we should not fear having the code with a BSD license on it. Microsoft will still get the technology behind it. Besides, bragging rights always seem to outway using the best code, and Microsoft does enjoy that use that right to its fullest.

What is journaling? by GuySmiley · 1999-08-11 23:00 · Score: 1

Pardon my ignorance, but how does journaling work? I hear this new buzz word a lot, but have not found a simple explanation.

--
Hey, leave comments about my mother out of this!

Who cares? by Anonymous Coward · 1999-08-11 23:00 · Score: 0

(a) Who cares?
(b) You can't integrate other Linux kernel enhancments into BSD
(c) SGI is getting into Linux. It doesn't give a damn about BSD (as most of us don't)
(d) Freer license would mean it could get into NT as well
(e) If BSD really wants it, BSD can GPL itself

Chew on this D.H. Brown by microbob · 1999-08-11 23:01 · Score: 2

Chew on this D.H. Brown.

'As for "keeping a log", that is a similar over-simplification. The reference is to "journal file system" a.k.a. a logging file system. This will probably get fixed in Linux within six months, but is currently a definite negative (long fsck times for large storage systems) and just one example of the many criteria we examined in our report.'

I suppose Linux will move up a notch?

It will in my book.

Jim

geez, enough complaining by cpeikert · 1999-08-11 23:02 · Score: 2

Gosh, all I see is whining about how this is "old news" - I don't recall ever seeing exactly what license XFS was going to be distributed under.

Regardless, let's showing a little fscking gratitude (forgive the pun) to SGI! They're supporting Linux on their workstations, their developers have always contributed to the kernel, and now they're freeing a huge, important chunk of code that is desperately needed and would be a massive development effort to start from scratch! Hoorah for SGI!

So what if it's a pre-emptive strike against IBM's journaling file system? More free software is good free software! Now we have more choices, and people can pick what's right for their uses, and may the best product "win", because in the end, we're all going to win. Huzzah! :)

We take em to court by Ben+Smith · 1999-08-11 23:04 · Score: 1

True, the FSF doesn't have the resources to take every 'suspect' software product into court, especially large software corps., but your suspicions are all the more reason why you should donate to the fsf. Code can only free so much, you need hard cash to put the teeth behind the GPL.
-Ben

--
-Ben
bensmith@biz1.net

Slashdot Mirror

XFS to be released under the GPL

244 comments