Pierangelo Masarati wrote:

Howard Chu wrote:

...

ando@sys-net.it wrote:

...

If it works, then we still have an issue: operations for "o=1,o=2" will probably be intercepted by the back-bdb. However, I understand that since the first database is glued, it wants to know whom it's glued to.

A fix is in HEAD. It will allow this scenario to succeed, but it will also allow completely broken subordinate configs to be dynamically created in cn=config. (configs read from slapd.conf/slapd.d will still be error checked appropriately.) As such, it needs a bit more thought.

Need ldap-txn, don't we?

I guess that could help. We haven't figured out yet how to propagate TXNs in syncrepl though.