[csw-pkgsubmissions] newpkgs py_lxml

[Sebastian Kayser]

Philip Brown wrote on 02.03.2010 00:47:
> meanwhile, we still have to justify why exactly we are packaging
> up py_lxml, when py_libxml2 exists, and we need to do it
> in a way that is reflected in its description.
> 
> your original naming choice, was:
> 
> "lxml is a Pythonic binding for the libxml2 and libxslt libraries."
> 
> The thing is, most reasonable coders would reasonably presume, that
> ALL py_XYZ packages are going to be "Pythonic", unless informed
> otherwise.
> Most people will view "Python Bindings" and "Pythonic binding" as
> identical in meaning.
> 
> So if a user is looking at only the two lines,
> 
> py_libxml2 - XML Parser Library Python Bindings
> and
> py_lxml - Pythonic binding for the libxml2 and libxslt libraries
> 
> they are probably going to have the reaction of, "well, that's just
> dumb: what the #@$@ is the difference between the two?? The
> descriptions are basically the same thing! Both are
> [python language bindings for libxml2]. How useless."

Anyone who bases his judgment of software usefulness on the package
description and who isn't doing python programming in the XML field
might think that.

All the others, those who actually need these modules, will be happy to
find their python module of choice in the list. Let's not forget, it's
the software which primarily provides value to the user, not a
comprehensive, disambiguating one-line description.

> So, please pick a description that differentiates better between the two.

Here is a nice example regarding the relevance of descriptions: a list
of descriptions for the text browser packages that we ship. It's _the
very first time_ that I even looked at them [1,2,3,4].

  w3m    - Text-based web browser
  links  - Textbased browser with frames, ssl, and menus
  elinks - An advanced text mode web browser
  lynx   - text browser for the World Wide Web

To follow your argument here, we would have to say: "why do we e.g. ship
lynx?" It's not obvious from its description. How useless. And btw., you
just recently released an updated version of lynx with exactly this
description (not that I would care at all and I do apologize upfront to
the affected maintainers in case you should decide to pursue explanatory
descriptions for text browsers).

Now even worse, if I were to base my software judgment purely on the
description, I might get the impression that links is the only one with
SSL support (explicitly mentions it, while the others don't) which would
be a totally false assumption. Fortunately I am a sane person (except
for engaging in this discussion of course) and know better.

In other words: IMHO people couldn't care less about the package
description as long as the description isn't misleading and their distro
actually ships the software they are after.

Sebastian

P.S.: I came up with a re-written description now, which looks good to
me. Looking at it now, one might think that was easy, but for me as a
non-native speaker, during a week more than full of work it somehow
wasn't at all ... and being blocked on a package release by one single
friggin description line certainly feels __very__ disproportionate
within our manageable team of active maintainers.

[1] http://www.opencsw.org/packages/w3m
[2] http://www.opencsw.org/packages/links
[3] http://www.opencsw.org/packages/elinks
[4] http://www.opencsw.org/packages/lynx