I fail to see this. People have repositories to do builds. If they're doing builds with or without maven, they're copying around libraries. If you have space to build your war file you must have space for usually several copies of the libs involved.
Admittedly I'm still on beta-7 for the moment, and it downloads larger amounts.
Still the majority of these jars are exactly the same and it seems like overkill to make everyone keep a separate repo just so that one project they are playing with doesn't interfere with the others.
I'm not talking about working on shared components - I'm referring to "subprojects", for example the model code that goes inside a webapp.
Helping with reducing the development costs of working with versioned artifacts is what maven's all about. If you want an unstructured build process, a tool thats all about structured builds isn't going to fit your process very well.I'd hardly say its an unstructured build process. I appreciate all the structure its added, but I don't think that every checkout should require editing project xml to set pom.currentVersion to "bporter-dev-HEAD" or "bporter-dev-SOME_BRANCH". This is really the only solution I see to not only avoiding sharing a repository, but sharing a repository per project. Essentially what needs to happen in this case is that project.properties defines the local repo to be target/repo, and then reactor driven ones under it must do the same, using ../target/repo and so on.
Seriously, fighting against the 'best practices' that maven tries to enforce will more that likely mean that using it will /cost you more/ than whatever you're doing right now. Maybe maven isn't for you at this time?
I think you misinterpreted what I was trying to say (or more likely I did a poor job of saying it). And I've realised since that even if we versioned each release, it still doesn't solve the problem unless you put a developer & branch name on the version (or use developer/branch specific repositories).
I'm more than happy to figure use whatever disk space and get it if necessary, but to me, setting a repository into every CVS checkout doesn't sound like the right solution. I don't see why, in a reactor based build, the reactor can't use the built artifacts without having to install them.
Cheers, Brett
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
