[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: partial replication of entries/attributes

Howard Chu wrote:
> I am about to commit a "partial replication" patch to the HEAD.
> This patch adds a "attr=<attribute list>" keyword to the replica directive.
> If specified, then only attributes in the given list will appear in the
> replog
> for the given replica. If an operation occurs that doesn't affect any of the
> listed attributes, no log is generated.
> Since this list affects both adds and modifies, it would be worthwhile to
> ensure that "objectclass" is always on this list if it's used at all.
> Otherwise you could add an entry that is completely ignored, but later have
> a modify come along that references the entry. (The replication would fail
> since the slave never saw the add.)

Good job, although it's going to give a lot of headaches:
if one entry is stripped some attributes but all the values
of objectclass pass, then it may be broken; would you consider
this a problem (e.g. no replica even of those info that were 
allowed) or a feature (the test didn't pass so it was a success)?

Maybe we'd consider the automatic inclusion of objectClass among
the replicated attributes, and select only those values that 
comply with the listed attributes (I'm talking in principle,
I know the answer: that's what we can get now; in the future 
we'll do LCUP :).


Dr. Pierangelo Masarati               | voice: +39 02 2399 8309
Dip. Ing. Aerospaziale                | fax:   +39 02 2399 8334
Politecnico di Milano                 |
via La Masa 34, 20156 Milano, Italy   |