I just got back from the 2011 Linux Plumbers
Conference. Definitely glad I went. It's been a while since I've
done any serious hacking on the Linux kernel, so there was a lot of
learning on my part. I'm glad to see that there is still a lot of
momentum and excitement in the community. It's also nice to hear people
admit some of Linux's shortcomings, and more importantly, discuss ways
to remedy those shortcomings.
The energy was inspiring, and some of the conundrums brought up have
tweaked my interest. I thought I'd jot a few things down that I think
might be worth revisiting, and hopefully I will get time to look into
them in future posts:
- Universally unique PIDs to avoid conflicts when migrating processes
- At the Containers
and Cgroups track, one thing that caught my attention was the
problem of restarting or "thawing" a process whose PID conflicts with
another process. Could applying UUIDs
solve this problem?
- CGroups vs OpenVZ vs jail/chroot vs SElinux for application layer
security - A couple
of comments
on slashdot about virtualization lead me to ask (even before LPC),
how could we selectively limit what apps have access to? The permissions
model of Android is a good start, but not fine-grained or dynamic
enough, I think. What models and "virtualization" methods are available
for this, and which one would hit that sweet spot of secure enough while
being fast, easy to use and unobtrusive?
- ARM not having hackable or desktop systems; why developing on a
handheld device is not ideal - At the last
keynote (on Friday), there was talk that Linux needs to see an
explosion of development on ARM like it did on x86. The only problem is,
where do you get a commodity, unlocked, hackable ARM system? And even if
you do, it's probably going to be a handheld device, so how do you hack
on that?
I hope to delve deeper into these topics later; who knows, if I
manage to scrounge up enough time, I might even be able to test out some
solutions to them.
