On Sun, Mar 8, 2009 at 4:05 PM, Antony Blakey <[email protected]> wrote: > > Firstly, are we correct in our assumptions about 'the CouchDB product' with > universally consistent (i.e. not configurable) semantics being 'the plan'? > What are other people's thoughts/advice about these issues? >
I can't speak for anyone else, but yes I tend to think of CouchDB as a product, much like MySQL or Apache's httpd server. I also sometimes think of it as a protocol. Eg if someone were to write a bundle of Ruby scripts that wrap Tokyo Cabinet or Hadoop, such that they passed the CouchDB test-suite, I'd be inclined to call that a CouchDB implementation or maybe CouchDB-compatible. The key thing here is that it'd be able to pass as a CouchDB node, both to other CouchDB nodes during replication, and to HTTP clients. If one were to use couch_btree and mochiweb to make a JSON store that is very Couch-like but doesn't pass the CouchDB tests or replicate with CouchDB nodes, I'd consider it to be a different product, and I think it would be misleading to call it CouchDB. Chris -- Chris Anderson http://jchris.mfdz.com
