[nickzoic]

The count({condition}) one is a worry. I'm guessing it is slow in the case where it has to page the index in in order to count it. I wonder if it is still a problem where the index is used a lot anyway. A fix in MongoDB would seem a lot better solution than having everyone implement their own hacky count-caching solution.

EDIT: Actually, looking at the bug reports, sounds like maybe lock contention on the index?

The master/slave replication problem seems bad but I think it can be worked around (for my particular project) with a flag on the user session ... if they've performed a write in the last 30 seconds, set slaveOkay = false. Users who are just browsing may experience a slight delay in seeing new documents but users who are editing stuff will see their edits immediately.