[Python-Dev] "setuptools has divided the Python community"

[Stephen J. Turnbull]

David Cournapeau writes:

 > On Thu, Mar 26, 2009 at 12:02 PM, Nick Coghlan <ncoghlan at gmail.com> wrote:
 > 
 > > If that perception is accurate, then any changes likely need to focus on
 > > the *opposite* end of the toolchain: the part between the "<common
 > > packaging spec>" and the end users.
 > 
 > Yes - but is this part the job of python ?

Yes.  In the sense that what I'm hearing from this discussion is that
tools that are very satisfactory for the upstream end of the chain can
often result in formats that are inconvenient for some downstream
end.  Even if the downstream end doesn't use it, some of their users
may mix that into their own installation, and it's difficult for
anyone to sort the resulting mess.  Ie, Python may not be resposible
for that distribution or its format, but Python should take care not
to kick up sand that may blow into downstream's gearbox.

 > Why coming from eggs

Or other complex format?  Well, because downstream consumers will want
them.  Nick's point is that there are multiple consumers of Python
distributions.  So we have a toolchain

                           +-> E ------> downstream developer -+
                           |                                   |
                           |             +----------+          V
source -> build -> A -> B -+-> C -> D -> | END USER | <--------+
                           |             +----------+          A
                           |                                   |
                           +-> F -> OS distro -----------------+

I think the point is to push the B -> C or E or F split point as far
downstream as possible ... but no farther.  As I understand his
request for "focus on the opposite end", it's that.  He's not claiming
that a one-size-fits all linear distribution chain is possible.

Rather, he's asking that we ask that very question: is there an
existing format that would be a better place to split?  If so, then
the stdlib should be built on top of that format, and it should be a
standard part of Python.

But what we have so far is people are saying "of course if you do it
my way everything will work out fine", but then they're left waving
their hands and chanting spells when it comes to being concrete about
how their vaporware PMS will satisfy the needs of the other camps.  So
there's a need to analyze the requirements of downstream library
developers, workstation app developers, web developers, end users, and
OS distros (I hope I haven't left anybody out).