associate artifacts with their pom and checksums

timestamp checks should also be taken into consideration in this (ie, if the pom of an artifact has changed, the artifact is rediscovered)

consider also the case where a JAR is first discovered, then the POM arrives later or vice-versa (the care that needs to be taken is how the first one is treated if its companion is missing, and how the second operation can make it behave as if they'd been discovered together)

currently, the indexer deals with this in its own way. I'm now considering whether to do this at all, or whether to keep the discovery mechanism simple

while I'm now happy not to do this, I need to do a final review of the index design since it changed recently to ensure that separated discovery of poms and artifacts (esp. if they have a classifier) works correctly.

The process for discovery of artifacts is kept intentionally slim in archiva trunk.

All files are treated as artifacts. (pom files included).
So order of upload/deployment is irrelevant.

Artifacts are added to the database.
The browse screen uses the database.

When the database update occurs the pom files are read, resolved, merged, expanded, and saved into the database.

When a user browses into a specific project version, and the database update hasn't occured (yet) on the pom files for that, the database update occurs JIT, making the data available to the user.