RE: Is State-based LDUP needed?

Hi all,

Steven Legg wrote:

I think it boils down to two questions.

Q1: Do we insist that replication primitives are always sent in
CSN order per replica ID ?

Q2: Do we insist that superseded primitives are propagated anyway ?

My opinion on these two questions are:

Q1: I vote "yes" to this one.

Q2: I could go either way on this.  It would seem we would be "more 
correct, more of the time" if we insisted that superseded primitives are 
propagated - at the expense of a little more information being sent during 
replication sessions.  However, it would appear that LDUP will work either 
way (after all, "eventual consistency" is what is required).  Thus, I 
would allow this as an implementation choice - i.e. superseded primitives 
MAY (or MAY NOT) be propagated.  I also think that it is in the best 
interests of all involved if the update reconciliation procedures draft 
(or the update replication protocol draft) points out the issues involved 
if superseded primitives are NOT sent.  (It seems a shame that this 
analysis would only be documented in the mailing list archives).

Regards,
Tim Hahn

Internet: [email protected]
Internal: Timothy Hahn/Endicott/IBM@IBMUS or IBMUSM00(HAHNT)
phone: 607.752.6388     tie-line: 8/852.6388
fax: 607.752.3681