A solr-user list discussion led to some general thoughts about Solr that I think need some further discussion. I'm ready to open an issue, I just thought it might be better to define a direction first.

Some of the option/attribute names in config files aren't self-descriptive, or have become not quite correct due to Solr's evolution. One example is "instanceDir" but there are probably others.

I'm not sure that "instanceDir" was ever a good name. It probably made complete sense when multicore first came into being, as it was meant to replace multiple instances of Solr. Coming up with a better name is a little tricky. A simple and currently relevant replacement would be "coreDir" ... but if you're using SolrCloud, Jack Krupansky has put forth some names that might be better: replicaDir, shardReplicaDir, or even the wordy but extremely accurate collectionShardReplicaDir.

In recent years, Solr has evolved from multicore-capable to multicore in even the simple example. If we have a similar migration so that all Solr installations are SolrCloud installations, then having replicas instead of cores (replacing instanceDir with replicaDir) might be the right way to go. Will we ever have a larger abstraction than collections? If we think that could ever happen, we should probably think of a name for it.

I think that we need to start a general overhaul of various identifiers in config files and APIs, planning ahead to accommodate future (in)sanity. Because it could be extremely disruptive to ongoing development, that probably needs to happen in a branch.

Thanks,
Shawn

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to