Re: Question about (negative) caching

[ggood@netscape.com (Gordon Good)]

"Eric S. Johansson" wrote:

> On a related thought, it seems to me that one could use failed requests as
> a way to trigger replication before that record.  That is, if the record is
> not found, the slave server could query the master server to see if the
> master has that record.  If so, then the slave would update itself from the
> master.
>
> one advantage of this form of replication (replication by use) is that it
> makes it much easier for a slave server to be turned on and be brought up
> to date without freezing the master and replicating by bulk data copy.

That approach would only work when the LDAP client attempts to read a single
entry. In the event that the client performs a search that could potentially
return a number of entries, the slave would always need to just pass the query
on to the master, since it has no idea if it contains all the entries needed to
satisfy the client's search request.