A quick FAQ: the reason all DSA keys have been removed from fd.o and we aren't accepting any new ones is that they are vulnerable to man-in-the-middle attacks if they have ever been used (not just generated) on a system with a predictable RNG: see Steinar's summary of the maths. We're going with precedent of debian.org rejecting DSA keys, and a general desire to be safe rather than sorry. RSA keys are the default in OpenSSH anyway, so I'm not really sure why you'd want to generate DSA.

It's just not our week, I guess. For the past couple of days, gabe.freedesktop.org (www.fd.o, ns1.fd.o, lists.fd.o) has been playing up: it will just mysteriously die with no explanation. It appears to still be alive, but we can't access it at all. Bringing it back up yields nothing suspicious before it dies, so we're suspecting some bad hardware. This wouldn't be too major a problem, bar the loss of the above-named services and dri.fd.o, openclipart.org, and a few other services, if it weren't for other things conspiring against us.

All revision control services (git/CVS/SVN committing and checkout, web access to same), Bugzilla, xorg.fd.o, people.fd.o and planet.fd.o, are on separate machines which are unaffected by this failure. But, unfortunately, we got less than 24 hours notice that our secondary DNS server at MIT was being moved, and would be offline for a while. Two DNS servers gone. Our third DNS server, hosted in the UK, is still servicing requests just fine, but apparently our registrar never noticed the update. So the third DNS server according to the glue records is home.keithp.com, which is also off the air right now for some reason.

So, ns0.bluelinux.co.uk is still serving names, and you can still access some services if you use that, but we're not sure how long until normal DNS is back up, let alone how long it will take us to get gabe fully fixed.

Sorry for the outage, please bear with us. If you think it's frustrating not being able to check www.fd.o, try checking the remote console and watching /proc/meminfo hopelessly spew into the serial console as you are unable to log in ...

fd.o's CVS migration is finally complete. CVS has moved from gabe.freedesktop.org (formerly 'the freedesktop.org machine') to kemper.freedesktop.org, a new secure host. Anonymouse CVS and ViewCVS now run on annarchy.freedesktop.org.

This means that, technically, you should now be committing to:
username@cvs.freedesktop.org:/cvs/reponame
and checking out anonymously from:
:pserver:anoncvs@anoncvs.freedesktop.org:/cvs/reponame

Anything using '@freedesktop.org' or '@pdx.freedesktop.org' is incorrect and may, at some point in the future, stop working. Anything using '@cvs.freedesktop.org' is incorrect, but will probably keep working for the time being. The canonical URL for ViewCVS is now http://webcvs.freedesktop.org; cvs.freedesktop.org will probably keep issuing 302s, but that too may disappear at some point in the future, so update your URLs.

Some code has moved to GIT, too. I can't remember exactly what the process for committing there is, but the process for checking out is:
git-clone git://anongit.freedesktop.org/modulename

The pserver redirection to anoncvs.fd.o wasn't working for a while. I'll post what I did, and see if you can figure out why:
root@kemper:~% iptables -t nat -A PREROUTING -d kemper.freedesktop.org -p tcp --dport cvspserver -j DNAT --to-dest 131.252.208.36:2401

Just as a heads-up, gabe.fd.o, its primary MX, is now running postgrey. Some mails may take a while to get through as MXes back off too quickly, so don't be alarmed. If your mailer throws messages away after getting a 451, that's its fault. Remain seated. Please do not panic.

This is a public service announcement. If you have a mail server, and it is going to stay down for a month, and you have someone else forwarding mail for your entire domain -- to the tune of 15,000 mails a week -- for you, why don't you let them know? That way their machine doesn't fall over and die every time Postfix tries to look at the queue.

Mail isn't working out, so hopefully they read this. David Turner, please fix the onward relay host for freetype.org. PLEASE!

Aaron writes about the stupidity of DConf. To be honest, I haven't been following the DConf discussions very closely at all, so I can't really comment in any useful way on the proposed standard, the people pushing it, whatever. All I can say is that fd.o is currently two machines with a couple of people have root on, that happens to run some reasonably popular lists. Anyone can subscribe and post, so what one person says, or a few people say, on xdg, might not be reflective of the opinion of the people who will be implementing and integrating whichever technology finally comes about. I don't doubt that a useful technology will come out of this -- if DConf is a failure, so be it. I'm sure that if it is, then something useful will be borne of its ashes.

That being said, I think if any good standard comes about, it will be a very good thing. I think the goal is worthwhile enough to persist with any bumps that may come along the way. But then again, I'm just hot-airing on a couple of planets, not writing any code, so take my views with the appropriate grain of salt. :)

Linux/OSS Technical Project Manager (GNOME/fd.o).

Just looking at the title of that job ad makes me happy.

Don't do this, or this.

(OK, so it turns out that I'm not as original as I thought, and that someone actually posted this to that project's list yesterday.)

Make of this what you will:

01:03 < daniels> clee: how about we set a definite platform release date?
01:04 < clee> daniels: uh, do we know enough to do that at this point?
01:04 < daniels> clee: we haev time enough to find out
01:04 < daniels> clee: locked in the same room together
01:04 < daniels> well, house, but anyway
01:04 < daniels> and not locked as such
01:04 < clee> That's very true
01:05 < daniels> i'd like some notice, not just 'Surprise! A platform!'
01:05 < clee> lol
01:05 < daniels> how does july 31st sound?
01:05 < clee> dammit, you've uncovered my evil plot!
01:05 < clee> "SURPRISE! PLATFORM!"
01:05 < daniels> heh
01:05 < clee> subtitle: Slow Vengeance vol. 1
01:05 < daniels> slow is right :\
01:06 < clee> I think that July 31st sounds like it might actually be doable.
01:06 < daniels> it is so.

I have suspended all SSH access to fd.o (well, except my own), thanks to the nasty little DoS without a proper patch circulating (yes, I've seen the one on lkml, and I'm kind of hesitant to apply it). Sorry guys.

Phil Blundell fixed xtrans already, after like half an hour of looking at it; fresh eyes solved the problem rather than stubborn persistence. Word up, Phil.

That's how easy it is to fix incredibly annoying stuff that affects everyone, seriously. We've had Phil bust up xtrans and two others start to work on Xorg's loader (and thanks to Donnie for a massively helpful Gentoo bug report on weak symbols), but there doesn't seem to have been much other progress.

There is not a flood of volunteers now. You will not get caught up amongst all the other people falling over themselves to get involved in fd.o. So, come get involved, and I'll buy you beer some day. I think I owe Phil supplies enough until he passes out, for one.

freedesktop.org needs you. Every single one of you. We need hackers, documentors, translators, accountants, whatever. If you have too many hours in the day, or you're bored, or whatever, *please* consider helping freedesktop.org out.

The best way to get involved is with an individual project: I understand that D-BUS could do with some extra hacking love, and it definitely needs its documentation fleshed out. I don't know how HAL's going. There are lots of other projects. Modular X needs help.

Look through TODOs, do what needs to be done, and send the result to lists. Go through Bugzilla and fix what you can. The cool stuff coming out of freedesktop.org doesn't just happen as the result of an accident with a particle accelerator and a goat: it only happens when people hack on it.

There are four new faces in important positions at fd.o. In lexicographical order:

• Adam Conrad - admin. Adam is a 'dedicated' admin - it is the only task he currently performs at fd.o, and he has pledged to take ultimate responsibility for the machine. This should help keep it more secure and eliminate some of the more funky issues with it. Adam earns his food and shelter by being an admin, so hopefully he's alright.
• Chris Lee - release team member. Chris is the other member (along with myself) of the release team, and will work with me on the platform, et al.
• Colin Charles - listmaster. Colin's main work is in moderation of all the fd.o-operated lists, but also does other list-related tasks. This should cut turnaround time down immensely.
• Pascal Hakim - listmaster. Pasc is one of Debian's listmasters, and is currently mainly working on creative spam-reduction techniques, so Colin has to do less work.

WiFi from the iPaq is nifty, but the ability to IRC from lecture with PocketPuTTY is potentially dangerous (irssi users: '/bind u sb goto -10', '/bind d sb goto +10').

X work is progressing very smoothly, with the modular-infrastructure-in-monolithic-codebase server working with ATI acceleration, and the loader close to bearing fruit. We're finally here.

I plan to spend any free software time I have this weekend on the poor fledgling platform. That means articulating a coherent vision, and initial proposals for the first release. Maybe even finish the install-module migration. So, if you're around in the platform, be around this weekend. Help is always appreciated, hint hint.

So, ChipX86 finished up the styles for planet.fd.o, so I put them up, toyed with them a bit, added some new features, and wham. We're now cliche-of-the-month compatible. Word!