[Distutils] Builders vs Installers

[Daniel Holth]

On Wed, Mar 27, 2013 at 10:41 AM, Lennart Regebro <regebro at gmail.com> wrote:
> On Wed, Mar 27, 2013 at 3:16 PM, Paul Moore <p.f.moore at gmail.com> wrote:
>> But sometimes practicality beats purity. As an end user who wants to
>> just install packages, but who knows that not everything will be
>> available as wheels, I need to be able to build my own wheels.
>
> Can you explain to me why you as an end user can not just install the packages?
> Why do you need to first build wheels?
>
> //Lennart

It's because when you install lots of the same packages repeatedly you
might want it to be lightning fast the second time. The pip wheel
workflow also gives you a useful local copy of all the packages you
need, insulating yourself from pypi outages. This is the practical
side.

The long term / bigger picture use case is that the wheel format or an
equivalent manifest serves as a sort of packaging WSGI analogue -- a
static interface between builds and installs. We would remove the
"setup.py install" command entirely. In that world pip would have to
build the wheel because it couldn't "just install" the package.

The first convenient wheel tool was more like wheeler.py. It was just
a shell script that called pip install --no-install, and "setup.py
bdist_wheel for subdirectory in build directory".