On Tue, Aug 25, 2009 at 5:31 PM, David Goldenwrote:
On Tue, Aug 25, 2009 at 5:51 PM, Andreas J.
I like both some kind of autoregistration and keeping the process on
PAUSE. Autoregistration if done right could take off a lot of burdon
from pause admins. (I've started coding it once but failed to do it
right, so it stalled.) Keeping the DSLIP form on pause cares for a
centralized input validation.
If we go this direction, I'd like to argue for DSLIP in META, with
well-defined values that PAUSE validates on upload. Or at least D,
S & L. I don't really care about I (Interface) for META, and
'license' is already there for P.
This would have several benefits, I think. It would give a place to
keep DSL with distributions, which could be useful for static
analysis. Knowing what languages are needed could help the toolchain
automatically check for appropriate compilers. It could also be used
for eventually moving away from underscore in version numbers as a way
to signal PAUSE not to index a distribution.
PAUSE could parse META, validate it, and then update the database.
What do people think?
I think that sounds like a good idea in principal, though it's
somewhat orthogonal to solving the "abandoned distribution" problem.
If I've got abandoned distributions (and - what do you know - I do!)
I'm not likely to get my shit together enough to upload a new
distribution just for the purpose of changing flags in the META. But
I could easily visit PAUSE and push a couple buttons to change the
The question then becomes: what happens if someone uploads a META
containing a certain set of DSLIP flags, then changes them via the
PAUSE web interface? Is the META file supposed to get updated? (No,
because that would at the very least break any checksums/signatures.)
Should it stay out of sync? (Yuck.) Should PAUSE refuse to accept
DSLIP changes if DSLIP is set in the META? (Thus negating the whole
purpose we're talking about this...)
Also, I'm not sure it really makes sense to put "abandonedness"
information right in the distribution. When people are releasing
code, they rarely think they're creating an orphan. That kind of info
really sits one level above the distribution, e.g. if the code were
distributed on a user's web page, the user would put big red letters
saying "I'M NOT REALLY WORKING ON THIS CODE ANYMORE" next to the
download link, and they wouldn't touch the tarball itself. The only
difference for people hosting their packages on CPAN is that there's
no place to put those big red letters. Yet.