In the same way we support Elasticsearch as a backend, it would be nice to support Solr.

Solr is different from Elasticsearch: it introduces the notion of cores, which are separate indexes with separate configuration files, schemas and so on.

In a core, you can have one and only one schema which defines the way the data are stored. Note that the schema can be quite flexible as you have the notion of dynamic fields so you might decide that all *_int fields are mapped to an int and you can then add any field named with this scheme.

Cores are manageable through an API and you can also manage the underlying schema with the API.

AFAIK, it is possible to build join request across cores but there might be limitations: we would have to study this carefully if we want to support joins one day.

Solr 6 introduces a JSON request API which allows to build requests with JSON: https://cwiki.apache.org/confluence/display/solr/JSON+Request+API . It's really far more readable than using the "everything in the URL" API.