Re: fs-rep-sharing infinite loop (again)

Vlad Georgescu wrote:
> Can't we add a boolean flag to node-revision-id, set to TRUE if the node-rev-id
> reused a rep?

Nope, that's not quite right. A node-revision-id carries two different
rep-keys (one for file contents, one for properties), and they are
independently deltified. *BUT*. In the new forward-order world, we know
three things:

   - the only representations that we should be deltifying are those for the
     most-recently-committed transaction.
   - we needn't bother deltifying representations that weren't created in
     that transaction (because if they weren't, they were created in some
     prior transaction and should already have been deltified)
   - the schema already preserves for us the transactions in which a rep
     was created.

I'm almost finished with a patch that makes use of this knowledge right now.

-- 
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand