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

Re: moddn (ITS#1192)



At 06:23 AM 6/8/2001, Ralf Haferkamp wrote:
>Do you know how much effort it would be to implement subtree modrdn?

Well, it fairly significant.

You'd have to find all the entries in the subtree and lock them.
Then you'd have to see if there were any containing
knowledge information (aliases and referrals).  Depending on the
knowledge, the subtree may viewed as crossing DSAs boundaries
and not movable.  Assuming the subtree is self-contained, then
one needs to run through each entry and rename it (generating
new indices as needed).  Then inner knowledge (aliases within
the subtree referring to other entries in the subtree) should
be updated.  Other references should NOT be updated.  That
is, if some attribute such an 'owner' would be dangling due
to the rename, it should be left dangling.  Then unlock
everything.

That's all fairly straight forward, but remember that back-ldbm
does not use/support transactions between the server and the
datastore.  I wouldn't attempt to implement such a large
update without transaction support.  I'd suggest doing this in
back-bdb once it is further a long.

Kurt

>We
>probably need this feature here. 
>Can you give me some hints were the changes have to be done?
>
>
>On Thu, Jun 07, 2001 at 05:56:43PM +0000, Kurt D. Zeilenga wrote:
>> I am not aware of any plans to implement subtree rename
>> in back-ldbm.  Until it's implemented, returning a error
>> is the proper action.
>> 
>> Kurt
>> 
>> 
>
>-- 
>Regards,
>    Ralf