allow to use non-unique index ids

Unfortunately this is limitation in the indexer, the context id must match id in the index. All relevant code is actually in the nexus indexer component.

Can we not just put the id in the properties file sitting along side the URL. The user should simply have to point at the URL.

Jason, it is already done like this. You specify url and push refresh button, then id is retrieved.

So this should be closed then?

Rich, can you validate this or if it's part of your existing test suite close this up.

The id retrieved from repository is still used as a key (pretty same actually happens in Maven itself), if two repositories would specify the same id in their properties you won't be able to add them to the Maven Indexes view without manually changing the id.

Indeed, the behavior described still occurs.

Seems like the fix would be to warn the user that the ID already exists?

we should simply change the id to be the url retrieved from, it's the ony way to have it be unique. Warning the user it aready exists does nothing to help them since it would be out of their control to rename the remote index. We have the same problem in nexus and it handles it transparently by renaming the incoming id to avoid a conflict. m2e should do the same. As a user, i just want the index, i don't really care what the other guy chose to call it and if it conflicts with what some other guy called his.

Rich, it already warns user if id exists.

Brian, I don't completely agree with you about using url as an index id. First of all the idea is to be able to match Maven's repo ids (so we can implement the project-scoped indexes) and second - urls would produce duplicated index identities for the same index if mirrors are used.

I'd like to learn more about nexus algorithm for renaming ids automatically.

The id should be removed from the UI and the ID should be made to be unique by using a fully qualified domain name. The URL may change over time, but the user should not select the id, it should come from the index that is generated and be guaranteed to be unique.

I may be missing something, but why do we force users to manually enter indexes in the first place? I believe in majority of cases users really want indexes for repositories defined in their settings.xml and/or referenced by their projects. So should not m2e just download indexes of all referenced repositories automatically?

Jason, I suppose part about making index id fully qualified by domain name should be handled by something like Nexus. Right now two default Nexus installations are producing indexes with the same ids and I don't really see how these indexes can be distinguished automatically on the client side.

Igor, the above should answer your question why user is requested to enter the index id. Also, it is not necessary that index is always defined in settings.xml, e.g. central is usually not there, but I've seen some users declare their own proprietary repository with id "central".

There is MNGECLIPSE-725 about project-specific indexes.

when you get an index, just preprocess it to change the id silently based on where it came from.