[SciPy-dev] GSoC weekly report

dmitrey openopt@ukr....
Sat Jul 7 01:34:21 CDT 2007


Alan G Isaac wrote:

> This is drifting OT, but I will add one comment, since many students may face such situations.
>
> You are mixing together a few different issues.
> 1. What will you spend your time on.
> 2. How will it benefit you.
> 3. How will it benefit them.
> 4. Will they pay you.
>
>
> Cheers,
> Alan Isaac
>   
However, they (especially chief) doesn't believe in languages like 
python at all and don't respect them. In whole department only 2-3 
persons know something different from C/Fortran (afaik they had deal 
with MATLAB, in SolvOpt project).
Chief says "there were LOTS of languages, but Fortran is the most 
powerfull, it's faster than C++ in a factor of 5 or so and faster than C 
in a factor of 1.5-2! (however, I found these numbers too far from real, 
maybe the custom benchmark written in our department was too special). 
All serious organizations that I had contacted with use either Fortran 
or C/C++. And Fortran recently have implemented all features that are 
needed - garbage collection, object-oriented programming etc" (I guess 
he means f2003 or f95 standards). I had explained them that numpy has 
compiled C/fortran libraries as atlas, blas, lapack but still they don't 
believe in python (as interpreter language).
What about name, they said "so we must each time explain to everyone 
that OpenOpt and scikits.optimize is the same? How scikits.optimize 
users will know that scikits.optimize is same as OpenOpt? We see no 
benefits here."


>It is still the case that many businesses have trouble understanding how paying some staff work on free and open source software can be profitable to the company.  There a many ways this can work, and it is likely to be very firm specific.  

All our department software and other solutions are opensourse, and they 
perfectly understand opensourse community, licenses, etc.
But, as I have already mentioned, they doesn't accept Python seriously, 
as well as MATLAB. They have spent lots of years working with Fortran 
(or some with C/C++) and lots of them have some years left before 
retirement, so they certainly will not switch to any other lang.
BTW it was THEY who proved me to remain in opensourse sector when I 
proposed them collaborating with TOMLAB (tomopt.com).
So, if I will mention anything about opensourse, they will just answer 
me "OK, here are some our (fortran/c) projects, they are fully 
opensourse - you can freely copy, spread, modify them, etc" (however, 
the salaries are significantly  smaller there, despite they have some 
grants from abroad).


If you have the dual goals
>    - would like to work with these people, and
>    - would like to work on FOSS not just as a hobby
>then they will need to understand how your FOSS work can be 
>good for them.

The work on scikits package can't give them anything, because I just 
implement in Python some code, that had already been written in Fortran 
(or C) by our dept or ASAI or other our dept patrners. And they have no 
intentions to switch to python from fortran. they say "who knows, maybe 
3-4 years later all your numpy/scipy/scikits will become unused (other 
language will appear, like fortress, or Python/scipy will be suppressed 
by Ruby/Rnum (ruby numerical library)), and several years will spent for 
nothing. Also, the type of work (implementing solvers and working on OO 
Kernel) will not yield you any scientific results, you do the same 
things that had been already done long time ago".

Regards, Dmitrey.



More information about the Scipy-dev mailing list