[FFmpeg-devel] GSoC 2009: FFmpeg is in
Thu Mar 19 12:44:17 CET 2009
On Thu, Mar 19, 2009 at 12:12:18PM +0100, Diego Biurrun wrote:
> On Wed, Mar 18, 2009 at 08:11:07PM -0700, Mike Melanson wrote:
> > I would like to point out that we don't have a whole lot of mentors or
> > projects signed up right now. If you have a good idea for a project,
> > please make it known on the wiki. If you are willing to mentor for it,
> > even better.
> During the 3 years that we have participated in SoC, 21 projects were
> worked on by students, but only 6 were finished in time:
> VC-1 Decoder (Kostya)
> RealVideo 4 Decoder (Kostya)
> Matroska Muxer (David Conrad)
> Nellymoser Encoder (Bartolomiej Wolowiec)
> ALAC Encoder (Jai Menon)
> MXF Muxer (Zhentan Feng)
> Kostya was already an experienced dev with plenty of implemented codecs
> under his belt when he worked on SoC and even there it is debatable if
> the RV40 decoder was really finished.
Nothing debatable. By September I had produced code that was able to decode
RV3 and RV4 but not correctly at all and I had to work for a long time to
make it suitable for SVN and viewer's eyes. Partially the blame is on
certain persons who failed to provide me specs (luckily, I could use debug
tool written by Mike).
> Bartolomiej had participated in
> SoC the year before and not finished E-AC-3 then.
> So in conclusion we can say that our SoC projects are too hard.
> Especially when the task is implementing decoders, apparently.
Some encoders too.
> Qualification tasks did improve the situation somewhat. The students
> that did them performed better than those during the first year.
> Nonetheless most of them did not reach the end goal, i.e. seeing their
> code in SVN.
> Something needs to change. I suggest making unfinished projects from
> past SoC events eligible projects this year.
> Google's guidelines explicitly state that projects can be repeated.
> The goal of SoC is to teach students how to work in a FOSS context.
> This can be accomplished perfectly by picking up abandoned projects.
> Furthermore it teaches the students to work their way through somebody
> else's code, a skill no taught at universities, but much more important
> than writing code from the ground up in most jobs.
> FFmpeg would also profit from this. We have had too much half-finished
> code languishing in the SoC repo for far too long. Especially glaring
> examples are AAC and AMR. I'm under the impression that they might have
> been finished much earlier if they had not been blocked by their
> inclusion in SoC.
> Big tasks that never reach mainline do not improve FFmpeg. Smaller tasks
> that reach mainline do.
Well, writing a large chunks of code that can be finished later usually helps.
It gives people something to finish and not writing from scratch.
More information about the ffmpeg-devel