I partly explained this some time ago on this ML. I'll c&p it quickly and hope it helps
> consider the _rev token as an opaque value necessary for the mvcc (I'd call > it mvcc-token, since "revision" is often misunderstood). The hash algorithm > may even change someday. AFAIK it contains the document itself and also its > history (at least parts of it). The algorithm ensures that the same document > with the same history result in the same _rev token even on two different > nodes. This behavior is required for conflict detection/resolution. Best Sebastian On 05.10.2010, at 04:19, Robert Sosinski wrote: > Hello, > > I just wanted to ask how CouchDB generates the string used for the > _rev. The first part (being the number) is pretty self explanatory. > However, I have not yet figured out how the second part that looks > like a MD5 digest is created. Would it be possible to > deterministically predict the _rev "digest" before saving the > document? > > Any clarification would help. > > Thanks, > > -Robert
