[clug] Request for Ideas/Suggestions: Fedora Scientific Spin

[Amit Saha]

Hello Andrew:

On 07/03/2011 09:31 PM, Andrew Janke wrote:
> Hi all (and Amit)
>
> First up: +1 to you for having the initiative to get something going
> by yourself.

Thanks! Its my itch and it looked like that it would be useful to other 
people as well.
>
> On Sun, Jul 3, 2011 at 21:16, Brad Hards<bradh at frogmouth.net>  wrote:
>> On Sunday 03 July 2011 13:31:45 Amit Saha wrote:
>>> It would be of great help if the interested ones among you suggest some
>>> generic tools for numerical computing/scientific research..
>
> I know I'm late to the party but my (humble) suggestion is that you
> dump Fedora and use something Debian based, there is already so much
> more effort on the Debian front in this area that the RH based stuff
> pales into obsecurity.

Whereas I agree with that, the problem I have often found in my limited 
experience with contributing to open source projects is that I am better 
off when I can start off something relatively new, or from scratch. I 
tend to get started easily and I have the scope for better contributions 
before I am exhausted away by the exercise of identifying a particular 
area to contribute to. Also, Fedora seemed a relatively nice place to 
explore the idea of such a spin, since they already have other spins 
such as "Electronic Lab", "Design suite", etc.

But yes like I was pointed out to earlier, I am definitely going to use 
"Debian Science " as a reference point for my efforts in fedora, for 
this spin and in general on the scientific front.


>
> On the Debian front, one that I use and contribute to is here:
>
>     http://neuro.debian.net/
>
> Yes, it's focussed on a very narrow field but is based upon all these projects:
>
>     http://wiki.debian.org/DebianScience

Very cool. I shall use them as a reference, definitely!

>
> Now back to suggestions of what is needed in "scientific
> distributions", I'll put in a vote for an easily
> configurable/pre-configured batch system such as
> gridengine/Condor/PBS. So many research areas now depend on linux
> clustering to do analysis and this is a difficult thing to install for
> a novice user.
>
> This would include all the work that first must be done to setup
> LDAP/NIS for authentication, NFS for cross mounting home directories,
> perhaps a private local subnet? etc etc.  Think of it as a "cluster in
> box".
>
> That's just my suggestion though, and again is scratching my own itch.
>
> Good luck with it!

Thanks much for the suggestions and comments.


Best,
-Amit
-- 
http://echorand.me