Always number index names?

[lukavdplas]

At the moment, we use numbered index names (e.g. ianalyzer-times-1) in production settings, but locally, we typically don't number index versions. (e.g. ianalyzer-times).

This is mostly because keeping track of versions, or storing multiple versions concurrently, isn't important in a development setting. Also, if you're developing a corpus with a sample of the dataset, it's not uncommon to re-do the index dozens of times, so the version numbers are less insightful.

However, it's not ideal that this layer of complexity is only added in production settings. Developers are less likely to notice issues with the versioning system if they're not using it in their local server.

So: perhaps we should remove the option for versionless index names? That is, you can still assign any existing index name to a corpus, but the indexing process should always number index versions.