RE: thoughts on replication requirements

[Alan Lloyd <Alan.Lloyd@xxxxxxxxxxxxxxxxxxxx>]

> -----Original Message-----
> From:	Jeff.Hodges@xxxxxxxxxxxx [SMTP:Jeff.Hodges@xxxxxxxxxxxx]
> Sent:	Thursday, April 23, 1998 8:14 PM
> To:	LDAP Replication
> Cc:	Jeff.Hodges@xxxxxxxxxxxx
> Subject:	thoughts on replication requirements
> 
> Here's some thoughts I have on our replication requirements and
> overall 
> approach given the present version of the reqs I-D, traffic on the
> ietf-ldup & 
> ldup-repl lists, and some review of some papers on this topic.
> 
> I feel Tim was correct in saying..
> 
> > The underlying problem here, and what's behind many of the issues
> John
> > is raising, seems to be that it's not clear what we are proposing to
> > standardize. 
> 
	snip to save transmission


	Lots of good thoughts and references. A couple of mine.

	I like to think that X.500 directories did not mention
"synchronisation" because the implementation and resolution of that is a
lower level data implementation issue that should be serviced by
realtime clocks, maybe database tools or just lots of glass fibre
between the DSAs. X.500 defines the object oriented abstract model for
directories with its high level protocols - No sync was not an omission
- its just things like performance, integrity, attribute index issues
and sync are usually left to the implementor so that competitive
advanatges can be obtained.


	So while many people can study this distributed integrity thing
- it really does come down to mechanisms - which may or may not scale
universally - do have a cost and a failure point. For our part - very
high replication consistency and reliability can be obtained via the
database - database replicator which because it is out of the RDB (sub
X.500 object level) its fast and direct. In addition we can DISP and the
X.500 level - and that could be equally as fast if the computational and
transmission resources are applied correctly. - Fast and Consistency do
depend on value judgement and product mechanisms.
	Standardising these will be (as said) a long haul. Because what
is good and what is bad?


	Directory level replication mechanims and the delivery of
services around master and replica data is a better problem to address -

	regards alan