On May 14, 2014, at 5:33 PM, M.-A. Lemburg <[email protected]> wrote:

> On 14.05.2014 22:56, Donald Stufft wrote:
>> On my phone so I can't respond to everything here but I just want to say I 
>> don't think a discussion where we can't challenge each other's conclusions 
>> isn't going to go anywhere. Hopefully we are adults and can handle 
>> disagreement. 
> 
> There's nothing wrong with disagreeing on conclusions and
> agreeing on this disagreement, but trying to shut someone down is
> not the right way forward.
> 
> What I'm trying to get back into the discussion is the view on
> PyPI as a community resource, which does not only need to address
> the needs of users of installers, but also those of developers who
> register their packages with it and make the whole thing come to
> life.
> 
> PyPI is used by people through the browser, it's used by people
> with installer tools such as pip, easy_install, zc.buildout, conda, etc.
> and it's used by developers using distutils, setuptools and
> other build tools that can talk to the PyPI API.
> 
> All of these people have needs and requirements, so we need to
> find ways to make most of them happy. In discussions on this list,
> I often get the feeling that the package authors are underrepresented,
> and that's why I try to add some of package author views to
> the discussion and also views as user of other tools than pip.
> 
>>> On May 14, 2014, at 4:26 PM, "M.-A. Lemburg" <[email protected]> wrote:
>>> 
>>> Noah, please reread the subject line and the message that started
>>> this thread. If we want to have a useful discussion, calling someone's
>>> conclusion incorrect is not helpful.
> 

Of course PyPI is a community resource and package authors are important. I
know that intimately since I publish several things through PyPI that I wrote
and I'm involved with several other projects. I suspect most people on this
list have published at least one package. If anything we are top heavy in
the viewpoints of authors with practically no representation from people who
just want to install some stuff.

I *know* that users have the expectation that installing something from PyPI
means they are downloading it from PyPI. I know this because they tell me this,
constantly. I've dealt with users every single day who are struggling to deal
with the concepts introduced in PEP 438. Whenever I explain to them that pip
will scan PyPI for links to places that are not PyPI they think that is just
crazy. It completely violates their expectations. The *only* people I have ever
seen *not* surprised by that, are people who happen to already know how
pip/setuptools/etc works. Most of those people that also think it's crazy that
they do that but they aren't surprised by it.

Framing the hosting of files on PyPI as an "extra" feature makes it appear to
be something added that isn't part of it's core competency. Nothing could be
further from the truth. For most people, authors and users, the fact that PyPI
host packages *is* what PyPI is. A mere 7% of projects that are registered
with PyPI have any reliance what so ever on externally hosted files. The vast
bulk of those are projects where the author forgot to upload a file or two
and external hosting isn't their primary hosting method.

That being said, nobody is trying to mandate that everyone must upload to PyPI.
I've put forward a PEP that proposes a way to resolve the problems with
reliability and implicitness of the current method of not hosting PyPI with
centralizing around the multiple index/URL support. There are changes to PyPI
in that proposal to increase the discover-ability of the additional indexes
however I ultimately think that it is the right path forward which brings PyPI
inline with what the vast bulk of people (authors and users) expect but still
preserves the ability for people who don't want to, to not host on PyPI as well
as reduce the number of ways that users have to be aware of for things not
being hosted on PyPI. It follows a pattern that many users are used to through
their dealings with OS packages and other languages, it allows them to
explicitly opt in to where they install packages from, and it allows a whole
bunch of UX issues in the current tools to just be eliminated.

Please go view that proposal (PEP 470, I posted it to distutils-sig earlier
today) and comment on it.

-----------------
Donald Stufft
PGP: 0x6E3CBCE93372DCFA // 7C6B 7C5D 5E2B 6356 A926 F04F 6E3C BCE9 3372 DCFA

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

_______________________________________________
Distutils-SIG maillist  -  [email protected]
https://mail.python.org/mailman/listinfo/distutils-sig

Reply via email to