[Distutils] Need for respect (was: PEP 438, pip and --allow-external)

Paul Moore p.f.moore at gmail.com
Thu May 15 14:53:52 CEST 2014


On 15 May 2014 12:38, Stefan Krah <stefan-usenet at bytereef.org> wrote:
>> While the opinions of core developers do matter, we're also far from being
>> representative of the wider Python community
>
> It's not only about core developers. The main point is that it's very hard to
> determine any general opinion of Python users. In fact, already in this thread
> we have four people stating that their expectations differ from the "official"
> ones.

This has always been a major difficulty with the PEP process, and any
similar consensus approach - the huge majority of users simply aren't
active in "the community". And furthermore, it's very hard to get
feedback from people who agree with you, so there's an inevitable bias
towards negative feedback - which isn't to say we can discount it! We
simply have to do the best we can.

For Windows users, having any sort of centralised package repository
is unusual, and massively better than the normal process for locating
applications. So I think it's fair to say that Windows users won't
have any specific expectations (unless they also have Linux/Unix
experience). That's the only platform I can talk about with any
confidence, but I *do* know that on Ubuntu users sometimes have to add
a non-standard PPA to install certain packages - that seems to me
quite similar to the multiple index model being proposed for pip/PyPI.
As a naive Ubuntu user, I "expect" apt-get install XXX to "just work",
but it doesn't bother me to find out that I need to add a new ppa (and
I have to find that out from docs - Ubuntu doesn't have the equivalent
of the discovery protocol included in the PEP).

It's absolutely true that I'm guessing here, but my view is that the
multi-index proposal is unlikely to be a major problem for end users
(particularly with the ability to add extra indexes to a global config
as is noted in the PEP). Certainly not compared with other issues such
as build/compiler issues and lack of binary wheels - the day I can
have "pip install pywin32 numpy scipy pyside" just work will be the
important day for me, and it won't matter one iota whether I need to
add a couple of --extra-index-url lines to a global config file to
make that work.

Paul (Speaking as an end user)


More information about the Distutils-SIG mailing list