[Soc-coordination] Project Idea fot GSoC
Daniel Pocock
daniel at pocock.com.au
Fri Mar 21 06:50:09 UTC 2014
Hi José,
The most urgent thing to do now is to insert your proposal in the Google
Melange site:
http://www.google-melange.com
It is just 12 hours until the deadline. If your proposal is not
registered on Melange at 19:00 UTC today, you can't participate.
The proposal should also be uploaded to the wiki as soon as possible but
for the purpose of the deadline, the Melange site is extremely strict
Regards,
Daniel
On 20/03/14 23:15, José Luis Sanroma Tato wrote:
> Hello again,
>
> I found a mentor and I have almost everything ready. Should the mentor
> creates the project in the wiki or should be myself?
>
> Regards,
> José Luis
>
>
> 2014-03-17 17:50 GMT+01:00 José Luis Sanroma Tato
> <josel.sanromatato at gmail.com <mailto:josel.sanromatato at gmail.com>>:
>
> Hi Daniel,
>
> Thanks for your comments. The item number 4 describes one of the
> uses of my project. :)
>
> I will write then in the debian-devel list.
>
> Regards,
> José Luis Sanroma
>
>
>
> 2014-03-17 14:52 GMT+01:00 Daniel Pocock <daniel at pocock.com.au
> <mailto:daniel at pocock.com.au>>:
>
>
> Hi José,
>
> Thanks for your email about this
>
> The most critical thing for you now is to find a mentor - if the
> Debian GSoC admins agree, the mentor could be somebody employed
> in your campus. Every project proposed to Google needs to have
> both a mentor and a student.
>
> On the problem and solution you describe:
>
> - one of my projects is remotely similar but not really the same
> thing (recursively building Java dependencies from source)
>
> - I've heard of big corporations using Ganglia (it is open
> source) as part of a strategy to find spare CPU cycles where
> they could run HPC tasks and also to measure the impact of those
> tasks on the machines concerned. You may be able to use Ganglia
> in a similar way to support your distributed builds. E.g. you
> could create a "logged_in_users" metric and when it goes down to
> 0, you use the person's machine.
>
> - there is some interest in automated rebuilds of full
> dependency hierarchies every time a dependency changes. The
> existing buildd servers may not be able to cope with that extra
> workload and a distributed solution may be helpful.
>
> - there are issues of trust when building official packages for
> the official Debian catalog - building on shared machines
> increases risk. Then again, given that DDs are trusted to build
> packages on their local machines, maybe spare clock cycles
> volunteered by DDs could also be trusted for rebuilds of
> official packages.
>
> These are all good topics for discussion on debian-devel too -
> you may also find a mentor there
>
> Regards,
>
> Daniel
>
>
>
> On 17/03/14 14:29, José Luis Sanroma Tato wrote:
>>
>> Hello,
>>
>> My name is José Luis Sanroma Tato and I am finishing my MSc in
>> computer Engineering. I heard about the Google Summer of Code
>> the last week.
>>
>> I don't know if it is possible to propose a project and this
>> is the reason I am writing to this mailing list.
>>
>> Currently I am working on my Master thesis that I expect to
>> present in June, and I think that maybe is a good starting
>> point to develop a bigger project because there will be still
>> problems to solve (adding architectures, issues with
>> repositories,...).
>>
>> I am working on a highly scalable and opportunistic
>> architecture to build Debian packages for different
>> architectures automatically taking care of the dependencies,
>> this project also takes part of the VIII Free Software
>> University Competition in Spain[1].
>>
>> At first appearance it looks like “buildd” but it's different
>> because it covers different needs.
>>
>> I am part of the ARCO research group[2] where we use Debian
>> and we have our own debian repository[3] where we build and
>> serve our debian packages. We use the computer of the workers
>> to build the packages. The problems that we have are:
>>
>> *
>>
>> We don't have a dedicated infrastructure to build software.
>>
>> *
>>
>> Furthermore, we don't know when a computer will be
>> available to build the packages due to the employees schedule.
>>
>> *
>>
>> We have some special needs, for instance, we usually work
>> with the Zeroc Ice middleware, which version 3.5[4] is not
>> part of the debian “stable” distribution and we need to
>> build it for “stable”.
>>
>>
>> My project tries to solves all these points by setting up a
>> distributed system in which each node is compound by computers
>> with some virtual machines as isolated and updated
>> environments. This isolated environments are used to build,
>> sign and upload the packages to the repository. I can go more
>> into details if you want.
>>
>> There will be some work to do, solving problems with
>> repositories, also adding more architectures (right now only
>> amd64 and i386 are supported)...
>>
>> I am not even a debian maintainer so I don't know if this
>> project could be useful in Debian somehow or if someone would
>> be interested in mentoring something like this to set new
>> objectives.
>>
>>
>> Regards,
>> José Luis Sanroma
>>
>> [1] http://www.concursosoftwarelibre.org/1314/proyectos/19
>>
>> [2] https://arco.esi.uclm.es/en
>>
>> [3] http://babel.esi.uclm.es/debian/
>>
>> [4] http://packages.qa.debian.org/z/zeroc-ice.html
>>
>>
>>
>> _______________________________________________
>> Soc-coordination mailing list
>> Soc-coordination at lists.alioth.debian.org <mailto:Soc-coordination at lists.alioth.debian.org>
>> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/soc-coordination
>
>
>
More information about the Soc-coordination
mailing list