dolfin team mailing list archive

["Johan Hoffman" <jhoffman@xxxxxxxxxx>]

> On Sat, Apr 21, 2007 at 07:42:39PM +0200, Garth N. Wells wrote:
>>
>>
>> Anders Logg wrote:
>> > Could we remove the modules (solvers)?
>> >
>> > We have discussed putting the modules in a separate repository for a
>> > while and we might as well do it now.
>> >
>> > I suggest we remove them now and then someone willing to maintain the
>> > modules in a separate project can make sure that the modules work
>> > against the new interface.
>> >
>> > Any objections? (some expected ;-)
>> >
>>
>> Given that most of the modules are not being regularly maintained and
>> have not gone beyond the embryonic stage, I'm inclined to say remove
>> them. Also, the demos have increased in sophistication and demonstrate
>> DOLFIN use for new users.

Well, this is not entirely true. That may be the case of the very basic
modules such as Poisson and the Heat equation. I guess the idea of these
modules are to be sort of a demo or light introduction to using dolfin for
developing one's own modules.

The interesting modules as far as I am concerned are the advanced modules
that implement recent and ongoing research, such as the Navier-Stokes
module for turbulent incompressible and compressible flow, and the
flud-structure interaction solver (maybe there are more), all being the
implementation of new computational methods for advanced applications.
These modules are well maintained and up to date, and curerently available
in dolfin-fsi.

> ok. If I don't hear anything else by Monday afternoon, I'll remove the
> modules.

Yes, maybe this is the natural thing, although I'd say think that it may
be useful to keep one or two simple modules for demo and testing of new
functionality in the dolfin kernel. Although, a simpler structure would be
to put these demo-solver under src/demo, and remove src/modules.

Then these simple modules could have the double role of demos and test for
new functionality. Or alternatively keep them under src/test.

>> In terms of maintenance, most modules are even less likely to be
>> maintained if they are spun off into a separate package. What about just
>> letting individuals make modules available? Module developers closely
>> involved with FEniCS project could publish their modules under
>> http://www.fenics.org/dev/, and others could place them on their own web
>> pages, with links from www.fenics.org.

To publish ones own modules on ones homepage is of course available for
anyone, and should be encouraged.

For our group, if the modules are taken out of dolfin I lean towards
publish our flow/fsi modules as a new fenics project under www.fenics.org,
focused on developing a generalized ALE solver for turbulent
incompressible/compressible fluid-structure interaction. These modules are
videly used at KTH and are very well maintained and are developing
rapidly, so this would be a FEniCS project that I expect would generate a
lot of interest and participation also outside KTH, which I got strong
indications of when this was last brought to discussion last fall.

>> Garth
>
> I'd prefer to have a maintained and polished set of solvers with
> somewhat similar interfaces all available as a single package
> dolfin-modules. But this would require that someone is willing to put
> in the work to maintain and develop the modules (and I'm not).

This might be useful for testing and demo of dolfin, so I think these
(small) set of solvers should rather belong to dolfin. The more
interesting modules are the ones that take computational mathematics
further, not just implements well known algorithms as dolfin modules.

> But if that's not the case, then the next best thing is to publish the
> modules separately under http://www.fenics.org/dev/.

This is possible, but I doubt they will generate much interest there.

> Anyway, I'll remove the modules and then we'll see what happens.

To remove the modules in dolfin-dev is fine by me, but I think we should
hear what is Johans J has to say before throwing them out (I happen to
know that he came back from a workshop in Barcelona this weekend).

/Johan




> /Anders
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev
>