Re: Excalibur/Cornerstone/James woes

[Aaron Knauf]

I have read and understood all of the previous posts in this thread. This whole conversation is only useful if we are agreed that we want to stay with avalon. If we don't, then someone need to say so before we waste any more time.
Assuming avalon stays, I can sum up my view by saying that my opinion remains pretty much unchanged from my previous post.

I have been stuck in various ruts before, where it was "too hard" to upgrade from a particular item of substandard software to a newer version. This has invariably been self inflicted (possibly for valid reasons). The james situation is the same.

We all agree that we do not want to release james with dependencies on HEAD revisions of anything. I'll go one further and say that we don't wnat to depend on an unsupported version of anything. (Not unless we are willing to take over maintenance of it ourselves, which means importing the source into our own repository.)
HEAD == unreproducible == unmaintainable == unsupportable == unprofessional.

I think we all agree that frequent stable releases of Avalon are required in order to make this work.

We all agree that gump is a good early warning system for detecting incompatibilities introduced between projects.
I think that there seems to be a misconception that a failed gump report requires an immediate fix. This is not the case. A failed gump report requires a decision to be made over the appropriate action necessary to resolve the problem. This may result in a discussion with the Avalon folks about making their change backward compatible, rather than an update of james. I expect that it would often result in a decision to make a note in a todo list for when the avalon change actually gets released.

In fact, I believe that we should be very reluctant to change james to accomodate changes in avalon cvs unless it is clear that a stable avalon release containing those changes will be available before the next james release. As others have mentioned, be don't what to block james releases.

I agree with Serge that the necessary updates to james should be made by the james community. I also agree with Danny that we don't want to all become avalon developers just to get james up to date.

In order to get james up to date initially, I still think that a branch would be useful. Sure, we will need to merge back into head when done. This can be made easier by merging head into the branch at frequent intervals. The reason for the branch is simply so that the avalon upgrade does not hinder current james development. Without the branch, you can either do it in head, or have just one developer do the whole lot.
Finally, I really don't think forking james over into the avalon repository is a good idea, because you just end up with an orphan that will be even harder to merge back in to james head. (Waste of time.)


ADK


--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>