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

Re: (ITS#3979) glue mixtures can confuse slapadd

Pierangelo Masarati wrote:
Howard Chu wrote:

Certainly there is no prescribed order, but we assume that a given implementation always uses the same order. Otherwise most of our test comparisons in the test suite are broken, and features like PagedResults cannot be implemented.

Back-meta, which is the reason that "async" appeared so early in slapo-glue(5), is a clear violation of the above... For that reason, in most of the back-meta tests, I need to use -S "" to ensure the same ordering in the results; as for the sizelimit, that's a completely different issue. I think you're absolutely right for a slapo-glue implementation that deals with data that is internally stored. In that case, the possibility to guarantee that the order is preserved in the implementation may be of value (can we change that order when changing versions? I think it already occurred at least once for backglue). But when dealing with data stored outside (as happens with the proxy backends) I wouldn't count much on the ordering. Even when plain back-ldap is used to proxy non-OpenLDAP software I think there's no way we can guarantee for the ordering of data returned from other vendor's implementations.

For now, I've dropped any mention of async from the subordinate description in slapd.conf(5).

 -- Howard Chu
 Chief Architect, Symas Corp.  http://www.symas.com
 Director, Highland Sun        http://highlandsun.com/hyc
 OpenLDAP Core Team            http://www.openldap.org/project/