archiva-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brett Porter <br...@apache.org>
Subject Re: Staging repositories (was: Re: GSoC projects?)
Date Wed, 28 Apr 2010 14:16:18 GMT

On 21/04/2010, at 2:33 AM, Eshan Sudharaka wrote:

> hi,
> I just added the gsoc proposal to the apache wiki(with out doing any
> change).
> https://cwiki.apache.org/confluence/display/ARCHIVA/Staging+Repository+Merge

I cleaned up the content to use wiki syntax and remove the GSOC specifics (that can all be
tracked separately - we just want to work on the technical proposal here). I haven't made
any changes to the content itself.

I'd like it if you were able to make most changes to the doc based on what we discuss here.
There's a few posts in the thread to pick up ideas from.

There's also plenty of questions to answer and adjust based on the thread so far:
- are the repositories temporary or permanent, and how do they get created?
- how will promotion work? what different options are there? (as Deng pointed out, the audit
logs are probably not the right approach here)
- how does a failed deployment get cleaned up?
- what permissions are needed exactly?

I addressed my thoughts in some previous messages. But don't feel like you have to agree with
me - the point here is to discuss alternatives and all convince each other until we find the
best solution.

I previously posted this: http://people.apache.org/~brett/staging-promotion.png. That has
an assumption of permanent staging repositories - which is an advantage that the URL is always
the same and doesn't require some finalisation to be active, but has a disadvantage that you
might mix releases. One way to handle that problem is to have artifact-level control of what
gets promoted (which you can make smart by reading the Maven metadata and finding out which
modules are related to each other). I discussed that more in the previous email.

One other thing to note: repositories are not necessarily Maven 2. Things like Maven metadata
merging should be handled by the repository API abstraction, rather than being baked into
the code. I'm happy to help discuss the design of this once we get past the "user level" ideas.

Cheers,
Brett

--
Brett Porter
brett@apache.org
http://brettporter.wordpress.com/





Mime
View raw message