[Savannah-register-public] [task #6160] Submission of OpenOpt

[Dmitrey]

Follow-up Comment #3, task #6160 (project administration):

Hi Stephan,

>How do you plan to use your Savannah account?
first of all I need subversion server somewhere, it's better than CVS from
sourceforge because allows to do atomic changes & renaming of files.

>However, we do not allow to host your project on Savannah and 
SourceForge at the same time, if Savannah is just a project mirror. 
Your project development should happen primarily on Savannah.

Ok, no problem.

>There is a companion site savannah.nongnu.org where we also host Free 
Software projects that are not part of the GNU Project, but run on 
free platforms.

I didn't understand what's the difference? Does hosting on gnu.org yields
more profits than in nongnu.org?

>A project name that says "open" will tend to lead people to think of the
project as "open source" instead of "free software". 
>We would be glad if you accept to use "free" instead of "open" in your
project name.
>While Open Source as defined by its founders means something pretty close to
Free Software, it's frequently misunderstood. For more information, please see
http://www.gnu.org/philosophy/free-....

Ok, so I visited your page, has read it carefully & find it to be very
essential.
On the other hand,
1) many people will treat word "free" to be not opensourse, just to be
something like shareware, like, for example, many instant messengers with
closed-code are. 
2) Unfortunately, word "FreeOpt" sounds very UGLY (at least in Ukrainian), so
some people, that could be interested in future in development and/or using
the tool, will laughing and avoiding this one.
3)I already registered the OpenOpt in
http://www.mathworks.com/matlabcentral/fileexchange/
-the main web area, where free m-code for MATLAB (and GNU Octave) are gained
Renaming of package is unavailable there. Unfortunately, no more then one
updating of package is allowed (according to the web area rules), so I had to
publish link, where my new versions with bugfixes & new features are
available, & it reduced my downloading rating - now openopt users doesn't use
mathworks link for download & it reduces my rate (btw I have 73 downloads for
now (last 16 days since reliase 1st ver; those numbers are for last 30
days)). 

4)Anyway, after reading the (brand)name of OpenOpt license ("GNU GPL") all
questions about free/open will gone away.

>We are careful about ethical issues and insist on producing software that is
not dependent on proprietary software.
So, now I have fixed some incompabilities with Octave & have OpenOpt version,
that runs in the GNU Octave environment (unfortunately, graphics output isn't
work in Octave yet, because they have some thoubles (some bugs) with
connecting MATLAB-style plot() to GNUPlot).
Anyway, area of commercial optimization packages is very big - lots of those
ones cost some thousand dollars. For example, TOMLAB/CONOPT costs 24500$
(http://tomopt.com/tomlab/products/prices/commercial.php), and it's only one
of the solver packages, that are included in the TOMLAB (3rd party commercial
extension optimization set of packages). So, Octave - GNU analog to MATLAB -
missed even fmincon() - the main MATLAB optimization routine! - which is used
(according to my observing MATLAB-related forums) aproximately in 30-40% of
optimization problems.
Of couse, it reduses Octave spreading speed very much, so I guess GNU should
be interested. BTW some our solvers, for example ralg, in some tasks are
better than any MATLAB or TOMLAB solver. Now I intend to connect RSQP toolbox
by Institute of Industrial Control, Zhejiang University
(http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=13046&objectType=file),
which also is better than some commercial solvers.
Of course, there are many free uncommercial optimization packages, but 
1) almost all of them are written in C/C++ or Fortran, so compilation
sometimes falls because of different OS & different compilers used; also RAD
(rapid application development) is unavailable there - it requires many time
to compile, to link, to catch bugs like segfault etc. 
2)Of course, in 99% cases Octave or Python are slower than C/Fortran, but
now, with implementing things like JIT accelerator (of already compiled,
running binary code) this difference turnes to be very small.
3)Almost all of the free optimization routines are written either as a
students' homework (hence the code is very poor) or as a accompanied software
to scientific articles or books, so it has unconvinient interface for
implementing into a project. Of course, graphic output is very poor or
absent.
4)Potential users of free software dislike using packages that are hosted at
links like www.msu.edu/optim/~researchers/bob_smith_79.html. Who knows - may
be some error will occure in future in already implemented project, and who
will fix the bugs, when link will become dead?
5)What about support: salaries in Ukraine are very small, so no much money
will be required from OpenOpt users for consulting & assisting them in
solving their problems.
6) more than half of them uses not 100% free licenses, and it requires to
study license every time you download something.

So, even adding "GNU " to the "OpenOpt" would turn customers to be much more
sure in the product.


    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/task/?6160>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/