[petsc-dev] PETSc-3.4 PR open at https://github.com/Homebrew/homebrew-science/pull/343

[Jed Brown]

Barry Smith <bsmith at mcs.anl.gov> writes:

>    They may be doing a decent job given the silly/archaic model they
>    have to work on, but after all these years maybe there is a need to
>    rethink the model. All these smart people wasting their time
>    working on complicated packaging systems to get around the limits
>    of the original model.

Managing dependencies is hard.  Linux distributions solve the problem
almost completely for normal software and normal users.  "aptitude
install octave" and you're on your way in seconds.  It will be
transparently upgraded along with everything else when requested,
security vulnerabilities in dependencies will be patched without needing
to reinstall the world, you can remove it (and any automatically-
installed packages) cleanly, you can link any collection of libraries
into another library or executable, etc.

Scientific software developers tend to put a great deal of effort into
making excuses for why they don't want to play in the ecosystem above.
It doesn't help that companies like Apple have no interest in helping,
and that companies like Cray and IBM are dedicated to breaking yet more
conventions to ensure that everyone knows they are special.

But if you want to make installation and distribution easy for the user,
especially if your software is "deep" in the stack, distribution
packaging is hard to beat.

>     Yes, Apple messed up the framework model (doesn't handle
>     documentation, not portable, doesn't handle versions well,
>     unneeded option names) but that doesn't mean one should stick with
>     a 40+ year old model just because it is the way it has been always
>     done. How about designing a better model?

http://www.xkcd.com/927/

>   Linux/Gnu has done a great deal of harm to computing because it
>   filled a partial vacuum by reinventing already 20 year old
>   technology preventing decent exploration of more modern innovative
>   approaches.  

The only way the GNU and Linux projects could have gotten started is by
being compatible.  Plan 9 is the awesome, elegant environment you
wanted.  At the end of the day, people settle for useful over elegant.

>   When a vacuum is filled with the same old stuff you can never know
>   what huge innovations you may lose out on.

Fine, let's not force our users to adopt our innovative directory
structure until we've established that it really and truly is better,
enough better to justify breaking all the tools that depend on the other
convention, and implemented a migration plan for everyone to move to the
new system.

In something as fundamental as dependency management, one capable but
mediocre system applied consistently has a better user experience than
an eternally-fragmented and mutually-incompatible system, even if each
member of the fragmented world is technically better than the mediocre
system.

I am in favor of strict adherence to these packaging conventions unless
we have a damn good reason not to, and any such reason must be something
of significant consequence to the user that is practically unworkable
without deviating from the convention.

By and large, our users want solvers that work well and are easy to call
and easy to adapt/extend to their applications; they don't want to think
about packaging.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20131012/4a372f7b/attachment.sig>