++ed by:

1 PAUSE user

Christine Spang
and 1 contributors


Return the replica's UUID


If the remote storage (RT) can not represent a whole changeset along with the prophet changeset uuid, then we need to create a seperate locally(?) stored map of: remote-subchangeset-identifier to changeset uuid. remote id to prophet record uuid

For each sync of the same remote source (RT), we need a unique prophet database domain.

if clkao syncs from RT, jesse can sync with clkao but not with RT directly with the same database.

Push to rt algorithm

apply a single changeset that's part of the push: - for each record in that changeset: - pull the record's txn list from the server - for each txn we don't know we've already seen, look at it - if it is from the changeset we just pushed, then store the id of the new transaction and originating uuid in the push-ticket store. - does that let us specify individual txns? or is it a high-water mark? - if it is _not_ from the changeset we just pushed, then do we just ignore it? how do we mark an out-of-order transaction as not-pulled?

Changesets we want to push from SD to RT and how they map

what do we do with cfs rt doesn't know about?