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

Re: lastmod and replication: slaves strip the attributes (ITS#110)

Moved discussion to -devel...

At 02:31 PM 3/24/99 GMT, az@Austria.EU.net wrote:
>in a replicated environment, the slave slapds strip the modif* and creat*
>attributes they get from the master, if configured without "lastmod"; 

That's basically acting as designed.

>(in theory they should add their _own_ attributes with "lastmod on" but i've 
>not been able to achieve this; anyways, nobody wants the slaves to maintain 
>their own versions of these attributes on the slaves...)

I agree that this would be desirable.  These particular operational
attributes probably only be modified by the master and then replicated
to the slave(s).  It would be interesting to see what the replication
drafts specify for handling of these (and other) operational attributes.

>idea for a fix: maybe add a "dontcare" setting to the lastmod attribute;
>slapd should then perceive modif* and creat* as regular attributes and 
>neither strip nor add these attributes on its own.

I think we need to come up with an approach to handle a wide variety
of operational attributes.  I'm thinking it would be feasible to
completely control such behavior through ACLs.