GSoC results!

[olafBuddenhagen]

Hi,

On Sun, Aug 31, 2008 at 06:54:15PM +0200, Arne Babenhauserheide wrote:

> Could you post an update with this years results to the GSoC project
> sites in the wiki? 

Well, I haven't touched the actual project pages (I guess the students
should update them, just like they did during the official GSoC); but I
now updated the main GSoC page (
http://www.bddebian.com/~wiki/community/gsoc/ ) and the project ideas
page ( http://www.bddebian.com/~wiki/community/gsoc/project_ideas/ ).
Would have done so before, but unfortunately I was quite swamped for the
past 2 1/2 weeks...

Here is the interesting part from the updated wiki page:

All in all we had five students working on a diverse selection of five
projects from our ideas list. All of the projects were more or less
successful:

- Sergiu Ivanov worked on namespace-based translator selection. Although
  he wasn't an official (sponsored) GSoC student, he worked on his
  project quite as steady as the other students (except for a two week
  vacation). The project however was hampered by various
  misunderstandings, wrong assumptions, and several major redesigns
  during the course of the work -- which is probably more our fault than
  the student's. In the end, he was able to complete nsmux (the main
  namespace proxy handling the magic filename lookups, running dynamic
  translators on demand); however, he still works on finishing the
  translator stack filtering, necessary to implement some of the desired
  functionality (accessing files while skipping existing translators).

- Zheng Da worked on network virtualization and some related topics. In
  spite of many open design question in the beginning, he did a lot of
  good work -- finishing not only the ethernet multiplexer and filter
  translators, which form the core of his project, but also a glibc
  patch to allow overriding the standard socket servers with environment
  variables; the devnode translator and a pfinet patch to allow accesing
  network devices through device files; support for settingthe network
  device in promiscuous mode in gnumach; a pfinet patch to use BPF for
  the packet filtering instead of the old Mach packet filters, and also
  to set a proper filter rule that really only passes the required
  packages to pfinet; a patch for the subhurd boot program to allow
  giving arbitrary virtual devices to the subhurd; and a proxy for the
  proc server, which allows running unmodified programs with a pseudo
  device master port instead of the real one -- providing some of the
  subhurd functionality without having to start a complete new system
  instance. He is still working on fixing some remaining issues, and on
  allowing subhurds to be run by normal users.

- Flavio Cruz was working on Lisp bindings for the Hurd interfaces, and
  did a great job: Not only he implemented bindings for all low-level
  interfaces as well as higher-level libraries for easy creation of
  translators and other hurdish programs, but also implemented a whole
  bunch of sample translators based on these bindings, some of them
  quite useful on their own account. He also fixed a few bugs in the
  Hurd he found along the way. Presently he is doing some further
  improvements, like additional abstractions and more sample
  translators.

- Andrei Barbu was working on porting a kernel instrumentation framework
  like dtrace or SystemTap. He implemented the necessary kernel
  infrastructure (and some nice general improvements along the way),
  making it possible to create tracing programs by hand; however, only
  at the end of the summer he realized that SystemTap is really
  extremely Linux-specific (while dtrace was ruled out already at the
  setout because of licensing problems), so there is no nice frontend
  yet. Unfortunately he was not able to continue work beyond the
  official deadline because of his PhD.

- Madhusudan.C.S was working on a new procfs implementation, to allow
  runnig existing programs based on Linux procfs out of the box. He
  managed to implement all the necessary information bits, so the most
  important procfs programs now work; and also fixed the procps program
  suite to actually build on the Hurd. There arestill some major bugs
  left, though. Aside from fixing the remaining bugs, he now works on
  adding some more information bits that are nontrivial to implement,
  and on fixing libgtop to work for us as well.

We decided to keep up the IRC meetings after the end of official GSoC,
so things can be properly wrapped up for upstream submission; but also
because the students want to continue discussing progress with their
ongoing work, problems, future directions etc.

I also think that regular IRC meetings are a good thing in general.

As always, the meetings are not only for (former) GSoC students and
mentors, but open to any interested party :-)

If someone of you is lurking on this list and would like to contribute,
but feel that you could do so better under formal mentoring: Please
speak up at the meeting! :-)

-antrik-