Comment 1 for bug 675590

The design Vincenzo and I came up with so far:

For each data migration that needs to happen, create a view that checks for the presence of the old feature, or the absence of the new feature on each document, and register a migration script or class in the code to deal with the documents that the view returns. Then on couchdb startup, run the migration views and fix any resulting documents.

We will need a way to register the views/migration scripts agains specific databases, or all databases. (Views should be smart enough that they ignore anything that is not a desktopcouch records, so that we don't mess with any non-desktopcouch data in the case we're talking to all dbs.)