Re: Updated document on kin&dyn API...

Herman Bruyninckx writes: 

> I've put an updated version of the kinematics and dynamics design
> document on the Orocos webpage:  
> 
>  <http://www.orocos.org/kindyn-doc.html> 
> 
> We really need to discuss APIs and coordinate representations.
> So, even more than ever, feedback is very welcome! 
> 
> (Not that I have any hope of getting feedback... This is the "open
>  source Cathedral" after all, isn't it. And even "release early"
> doesn't seem to generate much interest :-( )

I always read the documents with a certain care, but I don't
have time to follow all the things that come out on the web
on this subject, which BTW is only partially related to my work. 

The point is that, like I guess many colleagues, not only I do
research and develop software, I also need to use it to take
care of works for customers that are the justification for the
research and development I'm doing. 

As I've pointed out in a paper we're presenting at the
Multibody Dynamics 2003 conference in Lisboa, coming next week,
http://mbdyn.aero.polimi.it/~masarati/Publications/os2003.pdf
(where, btw, the OROCOS project is used as an example),
the key point of the "bazaar" is that OS projects need users
to create potential developers. OS projects never start from
scratch.  They start from an existing implementation which, even
if it's not optimal, creates a basin of users that, pushed by
needs, will develop it.  One critical point is that niche projects
do not attract users.  This is particularly true for multibody
software, shich is very specialistic: MBDyn sees some 2/3 downloads
per day (I see logs) but in one and a half year after release I got
feedback from 10/15 people, and only a third of them didn't give up
before compiling it.  I guess the software is so good that the other
900 downloaders were able to compile _and use_ it without problems ;) 

I expect robotics software to have a much wider basin of potential
users; RTAI definitely has, and, in fact, it is now mostly developed
outside of our department, with Paolo Mantegazza mostly doing
research and coordination.  However, both projects de facto address
niche problems, so ddo not expect much feedback unless you provide
from the beginning something that works well enough to attract users. 

I can tell you MBDyn works, and works well enough that I, as well
as other people, have been using it since mid '90s to do research
and consulting for aerospace companies.  Nevertheless, it's not
really attracting users, yet. 

I also participate in a totally different project, OpenLDAP.  That's
attracting users: it comes with every Linux release, and it's been
ported to all platforms, and I guess it's authenticating and profiling
more than 50% of the internet users. 

I suspect it is only a matter of critical mass. 

This is my 2 eurocents, based on real experience. 

p. 

Dr. Pierangelo Masarati               | voice: +39 02 2399 8309
Dip. Ing. Aerospaziale                | fax:   +39 02 2399 8334
Politecnico di Milano                 | mailto:[email protected]
via La Masa 34, 20156 Milano, Italy   | http://www.aero.polimi.it/~masarati