Hi,
here is a reviewed version of the Rename operation, as I know have a
better vision of what it implies.
Subentry rename :
Renaming a subentry has *no* impact on the DIt, but we have to update
the Subentry cache, as we store the CN in the subentry class. To
retrieve the modified subentry in the SubentryCache, we will use its DN
as an index. The 4 APCaches won't be impacted.
AP rename :
Here, we have to handle two different cases for each role. We will
process each role accordingly :
1) the AP is a SAP or an IAP for the processed role
If the AP is a SAP, we can safely rename it, rename all the
underlying entries, and we are done. We don't need to update the
seqNumber because we haven't modified anything that could impact the
subtree specification for any of the associated subentries. Enough said
that the subtree specification will only contain LocalName for the base,
chopBefore and chopAfter elements, all those names being relative to the
AP's DN.
If the AP is a IAP, it's a different story. We now have to check if
the chain of IAP for this role up to the parent SAP for this role have
subtree Specification that define a base, a chopBefore or chopAfter
localName, because if they do, then the evaluation will have to be done
again.
Here, we consider that renaming an AP should *not* impact the
subtreeSpecifications. It's up to the admin to modify those subtree
specification accordingly to the renaming being done.
If any subtree have been impacted, then we must update the associated
AP's seqNumber, in order to have the entries being reevaluated when
accessed. It's enough to get a new seqNumber and to inject it in the
modified IAP/SAP. We will do that for each IAP up to the parent SAP.
2) the AP is not a SAP nor an IAP for the processed role
Then the AP will be considered as a standard entry, and will be
processed accordingly (see below)
Entry rename :
We fall in the same case as when the AP is renamed, and it's an IAP. We
have to check the entry's parent AP, for each role, and check if the
associated subtree specification contain references to the previous
name. if so, we have to update the AP's seqNumber. If the entry's parent
AP is an IAP, we must continue until we get the parent SAP.
Note : this is the ultimate solution, but we can also play medieval on
APs, not checking the subentries. However, that would be brutal, because
we may impact may entries that are not under the modified entry. This is
typically a case where we may modify more entries than what we could if
we were to update the underlying entries immediately.
--
Regards,
Cordialement,
Emmanuel Lécharny
www.iktek.com
