[repo-coord] Re: Repo information inside rpm?

[Morten Kjeldgaard]

Axel Thimm wrote:

> In general I'd suggest any information that is required before of
> after the rpm builds (choosing platforms to build on, choosing repos
> to be shoved in), should be external to the specfile/rpm. Perhaps
> you'd like to propose a standard on such an external file format?

I suppose it would be possible, if each build recieved a unique id.
I could envision the following:

1. The user receives a unique id number from a centralized server 
somewhere (called "build-id" in the following).

2. The id is entered in the the "repo" string in the Distribution: field 
as discussed earlier, i.e. * .... id=452345251 ... *

3. The user builds the RPMs, in whichever way he/she prefers and is used 
to.

At this point, I am pretty vague about how to use the build-id in 
practice. But there are several possibilities, for example:

4. The user can run a program on the RPM which extracts information and 
sends it to the centralized server. This information can in principle be 
updated by the user, via a web-interface or some program.

But before going any further, let me then propose:

A) We agree on the Distribution: tag as the place to put the meta 
information. We use the keyworded form as proposed by Dag, since it allows 
the maximum amount of freedom: packagers kan employ their own keywords for 
their own purpose, and specific keywords for a standard purpose (for 
example the build-id).

B) Given A) the ideas on how to implement and use the build-id can simmer 
and cook in the community, and we can possibly try out different options 
in a test scenario.

Cheers,
Morten