Re: RFC: Keywords instead of Section

[Erik Steffl <steffl@bigfoot.com>]

Erich Schubert wrote:
...
> We should ask freshmeat + sourceforge if we could get a dump of their
> hierarchy structures (preferrably with some hit counts, so we can decide
> where to subdivide further etc.?)

  good idea, there are probably other classification systems as well.
maybe other systems would be a good inspiration (decimal system used by
libraries).

> What do you think that the first step should be?
> - Forming the keyword commitee, collecting keywords and starting to
>   classify packages (i.e. writing an overrides file for apt-ftparchive?)
> - do some user interface for this? (based on aptitude, which has already
>   some kind of support?)
> - pursue this discussion further and then contact debian-policy?

  1 define data structures first (to hold types and keywords) xml? some
db format? IMO xml is a way to go (human readable, there are standard
parsers etc.)

  2 write program(s) (or preferably (?) 2a API and 2b programs)

  3 create content (types and keywords)

  continue to do 3 as often as necessary. you might need to go back to 2
from time to time (rarely, bugfixes if nothing else (see a note below)).
You shouldn't have to go back to 1 ever (in reality you probably would
have to but it should be VERY rare).

  changes in time:

  if 1 changes: bad, you need to convert 3, and change 2
  if 2 changes: no big deal, 1 and 3 do not require change
  if 3 changes: this is supposed to happen, 1, 2 do not have to change

  NOTE: 2 might contain a lot of independent programs - e.g. you built
it into dselect, aptitude, apt-get, have some standalone implementations
etc...

	erik