Slashdot Mirror
Linux 3.9 Released
hypnosec writes "After a week's delay Linux 3.9 has finally been made available by Linus Torvalds. Last week Torvalds released the rc8 stating that he wasn't 'comfy' releasing the final version yet and that 'another week won't hurt.' Torvalds noted in this week's announcement that last week had been very quiet as there were not many commits and the ones which were there were 'really tiny' so he went ahead with the release of Linux 3.9."
it's not in TFS, not in TFA... what's new in 3.9?
Slashdot, fix the reply notifications... You won't get away with it...
Well that's how the progress bar counts so I don't see why it shouldn't be valid for everything else.
1... 2... many... 98%... 5... done
Among other things 3.9 includes experimental kernel-level Raid 5 and 6, caching slow of storage devices by fast ones, support for more graphics cards and audio devices, as well as improved power management.
What's wrong with that? That's how release candidates are supposed to work. You already do all your testing before the first rc. After releasing an rc you wait for anyone downstream to report problems. If there are no reports then that your final stage of QA is done and you accept the candidate.
http://www.h-online.com/open/features/What-s-new-in-Linux-3-9-1845705.html
09F91102 no, 455FE104 nope, F190A1E8 uh-uh, 7A5F8A09 that's not it, C87294CE no. Ah! 452F6E403CDF10714E41DFAA257D313F.
IMO just like Photoshop or Office, Linux kernel has been "ready" for a long time already. So every release can be just accounted to be "a little bit more of the same", some bug fixes, an odd device driver or the filesystem of the month. No big changes, you just get the latest refresh.
This version includes my patches for large HOSTS file support. Previously, the kernel used a fixed buffer size for the HOSTS file entries. relatively largeish (1 million+ entries) overflowed it and entries were dropped. 3.9 will dynamically resize (zero-copy algorithm) so the Linux KERNEL now officially has the best HOSTS file support anywhere.
You omitted the most epic part of the Windows family: Windows ME!
It's been around in the form of bcache for ages now and bcache is considered stable -- it is already in production-use. It's a pretty nifty thing, can be configured to your needs to quite a large degree and it's smart enough not to cache large, sequential reads/writes. If you're interested check out http://bcache.evilpiepirate.org/ and http://atlas.evilpiepirate.org/git/linux-bcache.git/tree/Documentation/bcache.txt?h=bcache-dev
Thanks for the link. I chased down the dm-cache stuff a bit, looking at it vs. bcache, which we'll see in June, probably. The primary differences seem to be that dm-cache is a bit more generic and easier to work with and it allows some manual allocation of the different types of disk data to be cached, while bcache seems to be targeting SSD's specifically, with wear-leveling sensitive write patterns, use of TRIM, and in-kernel code to validate the cache drives for preventing stupid user tricks.
EnhanceIO might also be targeting 3.10, which is a descendant of flashcache, which uses the layered device approach, while dm-cache and bcache are side-loaded. EnhanceIO uses udev rather than in-kernel code to prevent the stupid user tricks. I got the sense that 3.10 was possible for EnhanceIO but not as certain as bcache.
I'm current using flashcache on my non-ZFS systems; it looks like I'll be waiting for 3.10 and use bcache for my next upgrade.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Honestly I think there have been more interesting developments along the way in 3.x than there were from 2.4 -> 2.6.
I totally agree, but it's also fair to say that the efforts towards correctness from 2.4 to 2.6 have enabled the feature work happening in 3.x. I was just reading through the Xen commit about the feature to hot-plug new memory and CPU devices, and it's really so dependent on linux being able to say, "oh, you have a new memory module? OK, fine." which wasn't always the case.
And now for the flame-inducing observation: I think that through much of the 2000's, the BSD folks had a more mature and correct operating system to build on and that allowed them to add new features (e.g. pf, jails, zfs, dtrace) in an easier and faster way and we greatly benefited from that during that time period. Sure, linux had more drivers, but there was a fair bit of ugly kernel code to be faced. As of some point in the past couple years, linux has caught up, and now most of the really interesting new stuff (e.g. user namespaces) is probably going to be on linux, as that's where the momentum is. It'll probably be another three years before this becomes really obvious. Certainly there are still parts of linux with ugliness that need some TLC - I'm just talking about an inflection point, not an endpoint.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
What about Bob?
Im SAILING!!!!!
There is no right to feel safe thru security vaudeville at the expense of everyone's freedom, privacy and tax money.
Does it really matter that much to you whether its 3.9 or 3.8.1.24.96?
The number is incrementing at the same rate it used to, they just got rid of the stupid digit in the middle which ceased having any meaning years ago. Linux has been using the third digit as "minor" revision for years now, and they no longer do even-odd versioning with the second digit.
Seriously, everyone with a hangup about version numbers, get over yourself.
I kind of like the Ubuntu and MATLAB version numbering as it doesn't even pretend to be anything else than a simple date. I wonder if it could work great for projects like Linux or Firefox too -- projects that don't have clear major/minor cycles but which just keep on improving linearly. It has the bonus that you can see the release date quickly from the version.