Binary drivers

Amazing. Not that this exploit has been discovered. That’s no surprise at all. What’s amazing to me is that people will still continue to load large blocks of unknown, unauditable code into their kernel or as an xorg driver.

Guys, you are already potentially handing over control of your machine to someone who isn’t even willing to let you look at what they are doing with it. Now it turns out that what they are doing is writing poor code so that you are in fact handing over control to any sufficiently skilled attacker, and there is nothing you can do to fix it.

Actually, why am I bothering?

“Yes, but what about shiny games?” I imagine will be the response. I understand that for many this is a simple matter of priorities, but I just wish people would realise that when they hand over hard earned cash to a company, they are not just doing what is most convenient for themselves, but they are supporting that companys ideals and further empowering them. And that indirectly affects everyone.

libata PATA support

Well, it looks like libata PATA support is finally making it into the kernel. This is good news. Woo.

LRL06

I’ve got back from Lugradio Live 2006. I was an impressive event. Lots of talks, lots of people and a fair amount of drinking at the party.
On the “cool new software” side of things I found Mirco Müller’s Lowfat talk and demo intriguing. There is some good fun but also seriously important stuff going on there I think. Michael Dominik’s Diva talk useful. It looks very promising so far, although it did leave me wondering about why there is not more collaboration between diva and pitivi. I’m not suggesting they should merge or anything drastic like that; they clearly have different goals. I just think that they could share infrastructure like gnonlin. gdv really seems to fill the same role. Anyway, I will continue keeping a close eye on both projects.

Definitely the most impressive part of the whole weekend for me though was Simon Phipps, both in his “The zen of free” talk on Saturday, which was excellent and taught me a way of looking at open source that I didn’t even realise I was missing, and in his contribution to the “Mass debate” earlier in the day. Brilliant.

I’m sure many of the ones I missed were equally good. It’s just a shame we can’t see everything.

new Dragonfly release

It looks like DragonFly BSD is about to get a new release.

I think I’ll try it again once that is out. I like the DragonFly BSD design goals although I haven’t been keeping up on how close they are to being achieved. I also like the idea of massage passing in the kernel. Despite being a monolithic kernel it is starting to adopt microkernel features.

devfs

devfs is dead!

Yay! and thanks to Greg Kroah-Hartman for killing it.

GTK TreeModel implementations and language bindings.

I have tinkered with gtk+ many times in the past for various reasons. This has included my own mini-projects as well as changes to more widely used apps.

Only recently though have I tried anything more serious with it. This time it is via the gtk# binding using mono. Overally, I’m impressed with gtk#. It manages to stay close enough to the standard c api for gtk+ that anyone familiar with it could use gtk# easily, but also it gives a more c# friendly api where it matters.

One thing I wanted to do that is fairly important to my project was to write a custom TreeModel implementation. The reason is that the data will be retrieved on demand from a webservice running on another machine on the LAN and cached in the model implementation. I don’t want to use the usual TreeModel implementations because they use the “push model” and that would mean loading all the data ahead of time. Not good.

Easy, I thought. I just write a class that implemented TreeModel and code up the required methods. Not so. Apparently it is not possible yet to implement GInterfaces when using gtk#. Now, I have tried to think about what this might involve to implement in gtk# and I freely admit I got lost very quickly. Writing a binding for anything glib based cannot be a simple task.

However, I can’t help thinking this is something of a glaring omission. In fact, it pretty much means (as far as I can see) that I have to either:
1) Wait until gtk# allows this.
2) Put hacks into my project such as the ManagedTableModel that is floating around. (which will make portability slightly harder because I have to worry about DllImport - yes, I’m lazy)
3) Wait until gtk# provides something like ManagedTableModel itself
4) Choose another platform.

I started to wonder if this is possible in other gtk+ bindings. I haven’t searched exhaustively, but it appears that neither java-gnome or pygtk support this either. (anyone please correct me if I am wrong.) pygtk does have a solution for my case in the form of GenericTreeModel that can be extended and would work brilliantly.

Maybe I should be looking at using python and pygtk instead then?

Faster yum?

I just noticed that a new package called yum-metadata-parser has gone into rawhide. Interesting.

10 seconds googling later reveals that it’s a ~10 times faster yum metadata parser that is reported to also use a lot less memory. This might make fedora suddenly useable on a whole bunch of my machines. I look forward to trying it out.

Fix-it meet and apache weirdness

NLUG had quite a successful meeting last night at the Navigation. Those who brought along machines that they needed help with got help and I think everyone went home happy. It was good to see some old and new faces there. I took my laptop along although I didn’t have any specific problems that I needed fixing with it.

No problems, that is until I tried to run apache so that one of the other lug members could download some files from me. It ended up turning into the major unfixed problem of the evening. Many of us suspected it was just a bug somewhere (this laptop runs rawhide) and I eventually found that apache was refusing to serve static content where the files were 256 bytes or larger. Suddenly I suspected the kernel, and sendfile() in particular. Possibly related to recent splice() and tee() work. Once I configured apache to stop using sendfile() as a workaround all was well.

I guess if it’s not fixed in the next couple of kernel updates, I’ll try a vanilla git kernel before filing a bug somewhere.

Good Advocacy movie.

I have finally uploaded the (unedited) movie of Jono at his Linux Advocacy talk.
(and yes, I got the date wrong in the text but I can’t see how to change it).

Good Advocacy.

Last night we had a great talk from Jono Bacon (consultant, writer, and LUGRadio bloke) about good linux advocacy. I went to the talk expecting it to be an entertaining event with a good turnout, but not expecting to learn that much - being a good linux advocate is pretty much just common sense I thought. Well, it is actually. But Jono gave me some things to think about and make me look at the problem a bit differently.

Overall a great meet. Good beer at the Navigation as well. A few of us filmed the talk and I hope to have some appropriately encoded media files available fairly soon.

« Previous PageNext Page »