I have been using CouchDB to work out some proof of concept applications that interest me.  In that process, I have learn a good deal about the strength and weaknesses of CouchDB.  Here are my findings:

• Pros
  1. Massively scalable (no known limits with enough storage)
  2. Works well with Amazon EC2 and S3
  3. Uses REST and JSON
     Which means you can use any programming language or framework to write code for CouchDB.  There are already libraries for C#, Java, Ruby, Python, JavaScript, etc.
  4. Able to storage Base64 encoded blobs (PDFs, Txt, Doc, etc).
  5. It comes with a built in management interface called Futon
• Cons
  1. Horizontal scalability is not yet baked into CouchDB
  2. FullText indexing is not yet baked into CouchDB
  3. Schema-less data structures require a different set of disciplines from your typical programmers