Slashdot Mirror
Linux 3.5 Released
diegocg writes "Linux 3.5 has been released. New features include support for metadata checksums in Ext4, userspace probes for performance profiling with systemtap/perf, a simple sandboxing mechanism that can filter syscalls, a new network queue management algorithm designed to fight bufferbloat, support for checkpointing and restoring TCP connections, support for TCP Early Retransmit (RFC 5827), support for android-style opportunistic suspend, btrfs I/O failure statistics, and SCSI over Firewire and USB. Here's the full changelog."
It's funny. The Linux community put so much effort into trying to win the OS of the Desktop with so little success, but secretly won the battle of the OS on phones and tablets with hardly a fanboy.
Now this looks interesting. Hopefully it works as described on the net (http://lwn.net/Articles/479841/). Automatic suspend would be wonderful.
Ext4 metadata checksums. I like that. Note that it isn't data CRC checksums, just metadata. Still, I like the way Ext4 keeps evolving and getting tuned. Btrfs sounds really great, but it may still be some time before it is stable enough for my data storage needs.
Btrfs is stable enough for real data, if you run current releases (latest 3.4 or 3.5 kernel and btrfs-progs-19 current). I use it in both single drive systems and raid1 configurations with Fedora 17. Prior to converting the systems, I ran extensive failure testing (e.g., pulling power / data connection during active writes, system crashes, using a failing drive with media errors as part of a raid1, etc.) for about a month. I never lost a single byte of data in any test, confirmed by checksum scans on all data (against a backup) after each test cycle.
I actually trust btrfs now more than ext4 due to the ability to scrub the data and confirm integrity, which I do daily or weekly depending on the system.
Can You Say Linux? I Knew That You Could.
It depends on your definition of 'broke'. You don't have any of the functionality in the newer kernels (tens of thousands of patches to current), so if you want to use any of that it certainly is 'broke'. :)
Can You Say Linux? I Knew That You Could.
With the current implementation and just the 'autodefrag' option added to default, there is no perceptable difference in performance compared to ext4 for any of our machines, with any application. Recent testing at Phoronix (with 3.4) has btrfs getting closer to ext4 (running without lvm2 and md raid); I'm curious to see how its numbers look in 3.5. However, because btrfs integrates the functionality of lvm2 and md raid in a much more usable manner, as well as providing much more functionality, a small performance tradeoff would be acceptable (to me).
Can You Say Linux? I Knew That You Could.
That is fundamentally what politics is about: getting people who want different things to act together in a useful way.
So you're saying an orphaned inode ended up referenced by the wrong directory entry?
It's a way of transferring TCP connections to another server (along with transferring the IP address). That way, you could do hardware maintenance on the physical machine without losing anything because you've migrated it seamlessly to another machine. Of course, you have to transfer and restore everything else over too (running processes, memory, etc).
GNU has nothing to do with the kernel at all. The LiGnuX and then later gnu/linux renaming suggestions were for entire systems that contained software written as part of gnu projects and not a totally different project such as the linux kernel. The gnu kernel is called hurd.
The point of the renaming was stated to be to "advertise" gnu on the back of a higher profile project, but personally I think it was just petty MIT staffroom politics that escaped out into the world. "But what have you done lately Mr Stallman" turned into pretended ownership of linux which certainly has the above poster and a pile of journalists fooled.
It's part of the larger project for process/system checkpointing in general.
That is, saving the entire state of a process to storage such that it can start up again where it left off and not know the difference.
They use clustered file systems which have nothing to do with EXT/BTRFS/ZFS. If you're looking for a easy to manage, resilient, local filesystem, ZFS is about the only game in town.
Matt Blaze tweeted that Apple doesn't support the full resolution of the Retina display on the MacBook - the most you can set is 1920x1200, and it scales it from there. He also reports that there's a workaround which will let you get the full resolution.
But still, SRSLY? You'd think Apple could get font scaling correct, especially since they've been selling big desktop displays for years.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Back when I was running X Windows versions 10.x and early 11s, there was no requirement that I use TWM. And while the Sun 2 came with SunView, the Sun 3 could run either SunView or X, and you could get Grasshopper Group's implementation of NeWS if you preferred, which drove your screen in Postscript. Among other things, that meant that if you wanted to change the font size to match the size of your monitor and your eyesight, you just did it, and What You Saw Was What You Wanted. None of this "need a third-party developer's hack to use the full resolution of the expensive Retina Display you just bought" nonsense. But even if you were running X, you weren't limited to Motif or OpenLook; you could run whatever window manager you liked with it.
As far as "Ubuntu [does] not [have an SDK]" goes, you can use the Gnome SDK or KDE or LXDE or several other fairly full-featured SDKs.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
AFAICT, Linux-libre takes the standard Linux distribution and removes all software that doesn't have source code, most of which is device drivers, and also removes applications that don't have politically correct licenses. I'm not too worried about applications (apt-get easily fixes that), but I'd rather not load it on my hardware and find I don't have device drivers for the screen or the audio card or whatever. Does using Linux-Libre mean I can't use AMD graphics sets, or NVidia, or both? What about Intel chipsets?
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
some of us prefer blunt honesty over passive-aggressive politically correct doublespeak that dominates 'professional' interaction nowadays...those of us with spines, skin, and self-confidence anyway.
Linux, the kernel developed and distributed by Linus Torvalds et al, contains software that is included without source code, with obfuscated or obscured source code and code under non-Free Software licenses. Linux-libre removes these parts.
http://www.fsfla.org/svnwiki/selibre/linux-libre/
So, you're saying it is a lot, lot less functional, possibly even to the point of uselessness. Hmm. Doesn't sound like my cup of tea.
It's got nothing to do with politics, people with different preferences just happen to build different things.
Politics are about group preferences and group identity. It's got a heck of a lot to do with politics: There's no rational reason to offer standardized frameworks under an SDK for mobile, but avoid doing so for desktops. One of the irrational reasons behind the disparity is that the hacker and sysadmin culture view PCs as immediate kin to web servers hardware and regard any standardization or vertical integration in the stack that caters primarily to "luser" needs as a threat to their freedom and efficiency.
The overall 'distro' mindset causes each and every initiative to become hamstrung with idiotic assumptions, such as:
- Each application should be broken down into between 2 - 20 different pieces and scattered around one's hard drive.
- Changes for most system components are handled in exactly the same way and with the same priorities as high-level applications, and apps get to make dependency demands on the inner workings of the system. There is no clear distinction between system and apps for anything being updated, added or removed.
- Anything other than the kernel = "application", and this type of system-hacker nomenclature must be observed by everyone or they will be ridiculed as 'n00b'. The result is a kind of blindness to real issues that arise around interactions between apps and system.
- System coders > App coders, so we will just get Miguel and some of the ol' gang to whip up some applications that will put Microsoft and Adobe to shame (i.e. we'll draw from the pool of Linux system enthusiasts to write user-facing apps instead of creating a feature-stable environment with an SDK to attract both newbies, and experienced app coders who are only newbie to 'our' system). But the reality is that the particular hacker culture and general feature-instability act as a corrosive acid against the kind of userbase and developer community that a personal computer needs.
- More than 10 people like to manage their PC software within a paradigm designed for servers.
- Fewer GUI admin tools are better b/c people will just want to hit the CLI anyway. Avoid the GUI when describing solutions, even WRT office/productivity if possible.
- A myriad different admin tools for basic network connectivity are OK because people want 'choice' (esp. when they call up tech support for their ISP or application and the technician can't figure out what specific steps to tell the user).
- Each year, desktop users must learn to recognize "Linux" by the current and past iterations of the 4 or 5 desktop environments that are officially supported by each distro.
- App developers like to design their apps for a disembodied desktop environment, instead of viewing the OS layers underneath as equally accessible tools. They also like testing their app in several other desktop environments to ensure that it "plays well" with them.
- App devs love having to test and package on multiple distros, and they look forward to having many camps of distro maintainers telling them about app "bugs" that mean you have to help them fix the same issue in their systems over and over again for a number of years. They also love having maintainers pepper and berate them over wacky compile switches, setting defaults, patches, etc. and they way they like to refer to app devs as "upstream" instead of "author", as if "Linux" coding automatically entailed some sort of demotion.
- If one is an ISV (distributing a proprietary app as 3rd party), devs love being regarded as an oddball instead of the norm, and love being reminded constantly that so many of the compatibility issues with (untargetted) distros they keep having to read about could be automatically resolved if, gosh, the author would only release their app as open source so they could be merged with repository nirvana.
- App devs love hearing they should leave behind all the PC stuff and
I remember Microsoft tablets, there's no doubt they were first...
Except they weren't, maybe you mean "they came before". Microsoft Windows for Pen Computing was a ripoff of GO's PenPoint from 1991, right down the the notebook metaphor in their promotional slides. PenPoint launched on tablet computers from IBM and NCR before Microsoft was able to cobble together their first demo. And before GO there were attempts such as Pencept and Momenta to interact with an LCD screen with a tethered pen.
- and unusable
Stylus computing works great for annotating existing content (cue AT&T's "Some day you will fax from the beach" ads) but handwriting recognition remains terrible, even 98% successful recognition means you stop every few words to correct. PenPoint and software written for the O.S. won awards but never got traction, just like Windows for Pen Computing never went anywhere. A web full of content to consume didn't exist and there were only a few vertical markets of doctors, insurance adjusters, and construction supervisors to mark up existing documents (and fax them from the beach).
It was just like a PC, except with a stylus instead of a keyboard
Yes, Windows for Pen Computing just added an ink layer to the desktop. It was the other companies that rethought interaction: besides GO, the later Apple Newton and Palm PDA innovated. PenPoint provides a huge trove of prior art for direct screen manipulation, tap and drag, other gestures, and novel metaphors for a bookshelf, notebook, and page-turning, which is why Apple hasn't been able to broadly patent IOS features.
=S
It's funny, I've got a generic Dell PC I got for $300 from their outlet (Athlon etc etc.) Not a powerhouse, but not a shrinking violet either. I used my Mac Mini to download Debian Squeeze... and in 2 hours, with less interaction than a Windows install... I had a completely functional Linux desktop PC without doing anything in your list, except pick Debian. Back in college, I never got my et4000 card settings right for X on Slackware 0.99 (on floppies no less), but with screen and a familiarity with Amiga's CLI and DOS... I didn't miss it. (Plus that is the first time I got hooked on Nethack...)
Now, in the dim past just about everything was more difficult, to be sure... Getting games to run in DOS was also a magic trick. Then there was the myriad of other things that the CLI (which is where the computer originated) made easier for some, harder for others. Hell, Windows had a devil of a time keeping stable with the myriad of 3rd party drivers out there for Video and Sound cards alone.... Let's not forget NICs and so forth... And let's not diminish the fact that Windows used to be a graphical shell over DOS... for many years it was "hiding" DOS from the user...
Linux is a tool not everyone should use. There are idiots who shouldn't use a computer too. The fact that Linux has thrived in spite of Windows and Macintosh speaks more about the users and developers than it does about the drones who buy iPads and iPods because they're "hip". Those people don't use computers... they use appliances.
Here's a tip, though... if you hide everything from the user (a 'la original MacOS and returning to that I might add) it doesn't make them better at using a computer... it just makes them think all computers are magic. Which, in some people's case... I think should remain that way. :)
It's the Stay-Puft Marshmallow Man.
I've been using BTRFS in production since 3.3.1 with zero problems.
Before that, I did experience a partial fs meltdown on 3.2.x while stress testing a high number of snapshots with several million files/dirs and intense db activity. Then, the same test on 3.3.1 went flawlessly.
So I wouldn't recommend using BTRFS with anything below 3.3.1, but 3.4 or 3.5 should be fine.
Sorry Linux guys Bill Gates and IBM beat you too it many decades ago and that is the bad news.
I wasn't aware there had to be winners and losers in the game. Computing thrived with diversity, heck if it wasn't for diversity, we'd all be using some pablum invented by Bill Gates in his dorm... Thank goodness he was chasing rather than innovating... I shudder to think of the alternate universe that would've made.
Linux didn't "lose"... it's free. Linux is doing fine and several companies make a good bit of coin off it. I see Microsoft's "victory" as nothing more than eating a bit more of the pie than the other guys. Apple's got a huge market cap and tons of cash in the bank, yet they aren't even 20% of the PCs sold worldwide... So in the realm of "winners and losers"... we have to be a bit more objective, or at the very least, define what "win" means. Market share? Revenue? Mind share? Brand Loyalty? Whatever you pick, you end up with a different "winner." :) But I applaud Microsoft and Apple for really bullying everyone else for so long.. it makes me feel good that giant corporations want to actively screw me over and sell me the tools to do it with. I'm talking about both Apple and Microsoft in this regard.
As long as there are people who like to do it themselves... there will be Linux. Computing trends come and go, but revolutions stick around...
It's the Stay-Puft Marshmallow Man.
Linux pretty much kicks ass if you're an engineer though, because all of the things that make it horrible for end users give you enough access to pick your project apart down to assembly.
Yes, the patches (1,2) were merged back in May.
Can You Say Linux? I Knew That You Could.
This was a bug that was corrected (it was a problem in cache flushing). All my testing occured after the bug was addressed, and pulling drive data cables while actively writing, as well as pulling drive power cables, was part of my testing. No data loss occurred in any test. The btrfsck and btrfs scrub/balance were able to correct all errors that resulted following the drive recovery.
Can You Say Linux? I Knew That You Could.
Here's a tip, though... if you hide everything from the user (a 'la original MacOS and returning to that I might add)
They're not really hiding anything these days. They're mainly locking everything down, which is much worse. And fwiw, the original Mac OS didn't hide that much. You had to use ResEdit rather than emacs or vim to look at the internals of most stuff, but it's not like the average user knows what the /etc directory is, let alone what to change there (or even how to get sudo rights to do so, if the editor doesn't contain built-in functionality for that)
One big problem with classic Mac OS was that everything was way too accessible, resulting in people installing tons of extensions and control panels that changed the behaviour of the system, often resulting in instability. That was of course compounded by the lack of memory protection and pre-emptive multi-tasking, since every change to system behaviour affected everything, all the time. But a lack of use-accessible system customisation is definitely not how I'd ever describe the classic Mac OS.
Also, some other potential uses: http://criu.org/Usage_scenarios
ZFS seems like it could end up being a rather sad story - a very full featured, ahead of its time filesystem killed off by restrictive licensing.
Had ZFS been licensed under GPL, I suspect it would've been rolled part and parcel into Linux, and instead of losing Sun developers to btrfs, we'd have seen a completion of block pointer rewrite which would give ZFS the last basic features it really needs for end users (that is, ability to shrink filesystems online, reshape RAIDZ vdevs etc.)
It really is excellent - for a while I was running XFS+mdadm RAID6, and still managed to end up losing data because mdadm doesn't checksum and so would happily kick devices due to a bad sector here or there, and before you know it you're running with no redundancy and anything more it finds is a dead loss.
As it stands I suspect I'll roll on with ZFSonLinux until btrfs picks up something akin to RAIDZ3 and then migrate on over, but it really will be just a reinvention of the same featureset.
A word of warning, Linux Mint really needs to fix up how their installer drops X onto the system. The LiveCD will run great, but then you get all sorts of corruption till you get the right drivers installed.
It's a great distro - running it right now in fact - but oh god does that ever need to be fixed.
That's no longer the case.