Hi all,

Lex Spoon wrote:
> Craig Latta <craig at netjam.org> wrote:
>
>>>"Hey! Where did you find that? Oh, I didn't know about *that*
>>>server.... Hmmm, it isn't up now, do you have a copy you can email
>>>me?"
>>
>>	We could avoid that: having found any server in the network, a
user
>>would find out which other servers provide access to the desired
>>artifact.
>
>
> I don't suggest this for the universes themselves.  It sounds fine for
> the packages, but not for the *sets* of packages.
>
> A universe is supposed to be a collection of packages that are
> configured consistently and which have been tested with each other.
> Packages that are not in a universe are missing for a reason.  The
> problem with browsing around the net and offering the user whatever
> packages you see, is that you have no assurance that the packages are
> compatible.


Thank you, Lex,  for this definition of your understanding of the term
'universe'. The problem with this term 'universe' is that it is not a term
used in software engineering (*)

As we are here on a developer list mostly used by programmers which have
quite an elaborate knowledge of the field you can assume that
if you use more regular terms of the field you are better understood.

Why not strive for easier communication? I dare to
doubt that you want that because it seems you want to be 'innovative'.

The easiest thing to do that one superficial level has always been
to coin a new term - you have a lot of colleagues just doing the same
thing ;-)

A new term always can serve as a "catch-all-and-solve-all-
problems-magic-box". After some time people normally find out that
the concept has been here before in one way or the other and that
just plain work is needed to get things ahead.

The general connotation with the term 'universe' is that it is something
big, all-encompassing which is rather misleading in this case as we
actually speak of groups of packages which have been tested together.



> Everyone seems to assume I want regular users to have a list of 20
> different mini-universes that they merge together.  No.  I want the
> architecture to *allow* multiple universes.  I am picturing that most
> people would use either the standard unstable universe, or one of the
> stable release universes.  It's just icing that people can arrange
things
> in a lot of other ways should they desire.
>

This paragraph leads me to the conclusion that you speak of different
distributions.
We are back at the discussion of the various types of images
(basic, developer, kitchen-sink etc.)

Not bad, but why don't you relate to that  discussion?
I think it _is_ a good idea to bring this up again now and to come
up with different preconfigured sets of packages.

But for doing so a strategy of testing how the packages
which go in one distribution (universe as you like to call it)
work together?

The basic answer is pretty clear: Load all packages which one wants to go
in one distribution (universe) and do various tests (SUnit and others) to
check if nothing fails and that
they really work together. This is just hard work. If you have ideas how
to ease that work they are very much appreciated.

Diegos 'Full Assembly' is such a step on a factual level.
More efforts like this are needed (which can be implemented using existing
tools).

Cheers
Hannes



(*) Of course it is not forbidden to add new terms but a certain
justication for doing that is always appreciated. And the term should
be well chosen to catch on.