[ruby-talk:04411] Re: RAA development ideas (was: RE: Lookin g for inp ut on a 'links' page)

Me:
> > - We need testing and supporting community.
Hugh:
> Testing needs to be done, how this is done is another matter.
> Should there be peer review for submissions to the archive?

Peer review would be really good of course. I think, however, that it's more
part of the development and does not relate so much to the releases.
Naturally, ease of development (bug tracking, mailing lists, source code
versioning, documentation) has quite big priority too, but I was talking
about RAA as a software storage filled with releases of packages. The
development issues are very important but I feel something like Source Forge
is addressing these issues better than we will (maybe ever).

I see the role of RAA as a commonly know, unique, and consistently used
resource locator for Ruby related software. In essence entries is RAA are
collection of information about releases and where to download and find even
more information about them. So most of the time this information is
referring to author or project web pages and ftp servers. (Most of the time
these projects are ideally hosted by Source Forge or similar services, thus
the existence of the information is not tied to personal pages and ISP.)

Optionally RAA might "cache" releases, thus it's role is to act as a
collection of sources (and binaries, documents etc). In this case it would
be duplicating the information already on the project site; particularly
releases, maybe even docs, and the source code. This approach enables
effective automatic mirroring and advanced features like source browsing at
RAA, just like Source Forge or Gnome currently provide.

> Should the author(s) provide evidence of testing?

If they run the accompanied test suite, and attach it's output, that's
probably enough. The report

  <me> on <platform spec> with <software configuration> managed to build
  and test <package> with following notes: 
  <output from test suite>

is the evidence :).

I can't see any need to hand out challenges to testers. With challenges I'm
talking about some given input parameter set and requesting for processed
output. These would increase trust that the program if really working as
meant. But I can't imagine why people would like to "cheat" or forge
results. And if they want forge there's probably nothing preventing them, or
helping us to detect it.

	- Aleksi