fenics team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Nov 17, 2006 at 04:28:13AM -0600, Matthew Knepley wrote:
> On 11/17/06, Anders Logg <logg@xxxxxxxxx> wrote:
> > On Thu, Nov 16, 2006 at 11:26:50PM -0600, Andy Ray Terrel wrote:
> > > So people mentioned about the five minute attention span for building a
> > > system.  We did mention running a web service that might be able to
> > > help, such a feature is available through the Trilinos package but I
> > > haven't asked them how much it really is a bonus.
> >
> > This would be good to have. For example, people could input an FFC
> > form and get the generated code in return. This is ideal for a web
> > service since we have a simple model for what comes in and what comes
> > out. The problem would be that the FFC form is Python code, so the web
> > service would execute arbitrary Python code that someone enters into
> > the web form. So there are security issues.
> 
> This is not the big problem here I think (Python has an rshell). Whenever you
> distribute code (generated or not) you implicitly rely on a sahred
> body of support
> code. This is exactly the problem with Java. Its relied heavily on the libraries
> rather than just the machine. I cannot yet see how we could generate something
> that was usable apart from the FEniCS framework, in which case the user was
> able to install, so why have a webform at all?
> 
>    Matt

Good point. The code currently generated by FFC contains a large
number of #include <dolfin/...>. The next version of FFC will generate
UFC code that will not depend on anything other than the UFC
specification which is just one file (that could be thrown into the
generated code).

Still, the user would need to write the assembly-code so the support
code must be there anyway.

/Anders


> > > Another idea I have seen in growing popularity is to manage a wiki.  It
> > > would allow people to put up how things work more and give others the
> > > opportunity to look for more information.  I have actually never managed
> > > one of these things so I don't know how much work it would be.
> >
> > This would also be good to have, if someone is willing to maintain it.
> >
> > /Anders
> >
> >
> > > --Andy
> > >
> > > Anders Logg wrote:
> > >
> > > >Let me add a few items:
> > > >
> > > > - "Parallelization" of the mesh
> > > >
> > > > - Finish UFC specification 1.0
> > > >
> > > > - Move FFC to UFC
> > > >
> > > > - New build system
> > > >
> > > > - BDM, RT elements in FFC
> > > >
> > > > - DG in FFC (jump terms etc)
> > > >
> > > > - Design and implement UFL
> > > >
> > > > - Move FFC to UFL and integrate UFL in DOLFIN/PyDOLFIN
> > > >
> > > >/Anders
> > > >
> > > >
> > > >On Mon, Nov 13, 2006 at 03:49:26PM +0100, Johan Hoffman wrote:
> > > >
> > > >
> > > >>I agree with Johan; Garth was a top host! I had a great time, and I got
> > > >>back to Sweden without having been run over!
> > > >>
> > > >>Some things that came up, which we should pursue in one way or another:
> > > >>
> > > >>* Dolfin: surface integration (boundary + internal for discontinuous
> > > >>Galerkin etc.)
> > > >>* Dolfin mesh: hierarchical vs single mesh mode (geometric vs algebraic
> > > >>multigrid): support both.
> > > >>* Dolfin mesh: refinement/coarsening: reimplementation of broken
> > > >>Bey-algorihtm, test of "Matt-algorithm", test of simple split/collapse
> > > >>algorithms + smoothing/flipping.
> > > >>* DOLFIN: Full adaptive Navier-Stokes solver to be implemented (today
> > > >>primal solver + broken mesh refinement in place)
> > > >>* DOLFIN: Compressible flow module
> > > >>* Dolfin: ALE + free surface modules
> > > >>* Dolfin/FFC: automatic generation of error estimate/mesh indicators
> > > >>* Dolfin build system: Scons or Cmake...
> > > >>* Dolfin/FFC bc: assembly of strong bc directly in FFC (together with weak
> > > >>implementation)
> > > >>* dolfin-dev + dolfin-stable
> > > >>* Dolfin license: there seems to be an agreement on GPL -> LGPL
> > > >>
> > > >>/Johan
> > > >>
> > > >>
> > > >>
> > > >>
> > > >>>Hi all,
> > > >>>
> > > >>>The FEniCS'06 workshop was very successful: many interesting
> > > >>>discussions came up, problems were overcome and new plans were made. I
> > > >>>also had a good time personally, and it seemed like everyone else did
> > > >>>as well.
> > > >>>
> > > >>>I realize organizing a good conference is often unappreciated work.
> > > >>>But I think most people are aware that Garth provided the initial
> > > >>>conditions for the conference, so to speak, and had a hand in
> > > >>>essentially everything that happened. So in this case it's much
> > > >>>appreciated work, which I hope we'll be able to repay in kind in the
> > > >>>future.
> > > >>>
> > > >>>Garth mentioned at the end that it would be good to make a concrete
> > > >>>plan of all the things that came up. I think this is a good idea, and
> > > >>>it would be good if people could write down and post their plans while
> > > >>>it's still fresh.
> > > >>>
> > > >>>  Johan
> > > >>>_______________________________________________
> > > >>>FEniCS-dev mailing list
> > > >>>FEniCS-dev@xxxxxxxxxx
> > > >>>http://www.fenics.org/mailman/listinfo/fenics-dev
> > > >>>
> > > >>>
> > > >>>
> > > >>_______________________________________________
> > > >>FEniCS-dev mailing list
> > > >>FEniCS-dev@xxxxxxxxxx
> > > >>http://www.fenics.org/mailman/listinfo/fenics-dev
> > > >>
> > > >>
> > > >_______________________________________________
> > > >FEniCS-dev mailing list
> > > >FEniCS-dev@xxxxxxxxxx
> > > >http://www.fenics.org/mailman/listinfo/fenics-dev
> > > >
> > > >
> > >
> > _______________________________________________
> > FEniCS-dev mailing list
> > FEniCS-dev@xxxxxxxxxx
> > http://www.fenics.org/mailman/listinfo/fenics-dev
> >
> 
>