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

Re: "attrs=" directive in syncrepl

One of the problems here is that different attribute
lists have different restrictions.  Namely, the syncrepl
list is an LDAP attrslist (as it sent over the wire)
where ACL attrs are internal.  Also, exattrs is
(unnecessarily) inverted (since anlists support exclusions).
If syncrepl instead had attrs (to request) and saveattrs
(to save), we could gain symmetry and hence allow better

Anyways, I agree that cleanup and reuse is good here.


At 04:31 PM 9/26/2004, Howard Chu wrote:
>Pierangelo Masarati wrote:
>>I haven't followed the recent discussion about the "attrs=" directive too carefully,
>>so I don't recall, and I'm not able (or too lazy) to find out what was the final decision.
>>I note that by specifiying "attrs=*" one gets in fact "*,+" after a long way to the
>>si_anlist member of the syncrepl structure.  I was cleaning up the code for unrelated
>>issues, and I committed a fix that is still commented because I'm not sure if the
>>behavior was intended.  In fact, with my fix test019 fails; to avoid the failure, the
>>config files would need to explicitly have "attrs=*,+". Am I missing anything?
>I don't know, but this whole chunk of code needs to be simplified, as does the attrs= parser for slurpd replication. Right now we have
>   attrs=<list of included attrs>
>   attrs!=<list of excluded attrs>
>   attrs=!<excluded attr>,!<excluded attr>...
>   exattrs=<list of excluded attrs>
>Maintaining all of this is ridiculous. The anlist structure supports an exclude flag per item, we should be using it consistently.
>As such, there should only be one config syntax:
>   attrs=<attr>,<attr>,!<attr>,!<attr>...
> -- Howard Chu
> Chief Architect, Symas Corp.       Director, Highland Sun
> http://www.symas.com               http://highlandsun.com/hyc
> Symas: Premier OpenSource Development and Support