On Thu, 2010-12-02 at 17:49 -0500, Jeff Squyres wrote:
> On Dec 2, 2010, at 4:38 PM, Bernd Kallies wrote:
> >> 1.1 is pretty close to done. If you wanted to shift your work to be based on 1.1, I think you'd be pretty safe.
> > I'll try. Currently my wrapper implements only basic things, so there
> > should be no problem (because it is a wrapper, only). Problems would
> > arise when one wants to extend the number of implemented methods.
> > To be
> > honest, I expected some remarks about the completeness of the wrapper.
> My personal bias is always reading from the objects; I rarely use many of the accessors (simply because hardware may not be uniform). That's why I thought your first set of accessors was sufficient.
> My $0.02: those are ok. Go with that for a first version. Then get some real-world users and see what they ask for.
> >> Would you -- or your employer, if they own the code that you generate -- be able to sign this document?
> > The answer is yes.
> FWIW, Samuel made a good point to me off-list earlier today: keeping language bindings as a separate package is worthwhile because then they can use their own language-native build/install/packaging tools rather than have to deal with the GNU autotools. That's a relatively good argument to put stuff on CPAN (and/or push them via RPMs to the distros).
> At a bare minimum, I believe that if you host perl bindings on CPAN (and Red Hat Guy hosts python bindings elsewhere), we can definitely link to that site from our web site, README, ...etc.
So, in summary you say that I'd submit the perl-hwloc binding to CPAN.
I also believed that this should be better than distributing this within
hwloc, but I wanted to hear what you say first.
Dr. Bernd Kallies
Konrad-Zuse-Zentrum für Informationstechnik Berlin