Summary and call for discussion on text proposal

Johan Tibell johan.tibell at gmail.com
Sun Nov 7 13:11:41 EST 2010


On Sun, Nov 7, 2010 at 7:54 PM, Gregory Collins <greg at gregorycollins.net> wrote:
> Myself and several other people have been following this discussion
> with increasing levels of annoyance and frustration. My understanding
> was that the HP process was intended to help with the overall design
> of libraries and to head off serious problems before too much time is
> wasted on discussion, NOT to devolve into extended megathreads over
> which colour to paint the bike shed.

This was at least my intention when I co-authored the package addition
process document. I wanted to have a venue where experienced Haskell
hackers could help improve the quality of libraries that went into the
HP. I tried to model the process based on Python's PEP [1] process,
which seems have worked for them for the most part.

In my head, the process would help prevent bad design from getting
into the HP, where it might be harder to fix due to a wider user base
and higher stability expectations. To give an example: the creators of
the Python WSGI spec decided that the WSGI [2] implementation should
decode any URLs before passing them to the user. This turned out to be
a bad decision as URLs are no ambiguous. For example, these two URLs

    http://www.foo.com/hi%27there
    http://www.foo.com/hi/there

decode to the same URL and thus some information is lost, making the
WSGI API slightly less useful (i.e. for people who need to distinguish
between the two). I wanted a design review process that would help
catch, among other things, such corner cases that the original library
author might not have considered.

1. http://www.python.org/dev/peps/
2. http://wsgi.org/wsgi/

Johan


More information about the Libraries mailing list