Re: Glueing to back-dnssrv in 2.2

["Pierangelo Masarati" <ando@sys-net.it>]

[sorry for the empty reply earlier]
>
>>> I think the backport to 2.2, although not trivial, should be simple,
>>> because all you need to do is resolve the appropriate backend and call
>>> chk_referrals, if any.  If there's no consensus about officially
>>> backporting that feature, you'll have to live with a custom patch.
>>
>>There you go:
>><http://www.sys-net.it/~ando/Download/pierangelo.masarati-dnssrv+glue-re22.2005-01-28.patch>
>
> Thanks for the patch! Some comments:
>
>  - It appears to work for searches with an empty base DN, but fails
>    with noSuchObject if a base DN of a physical naming context is used.

I didn't check it enough, I guess; I couldn't get to configuring a DNS at
the moment, and I don't have any DNSSRV at hand (maybe I could use
OpenLDAP's as soon as I get on a faster connecton).

>
>  - Also, we would like some way to avoid databases glued to a database
> with
>    with the empty suffix from being removed from namingContexts in the
> root
>    DSE.

For that, you need to hack rootDSE info generation, which is
(correctly)designed to hide subordinte naming contexts; or (in HEAD/23)
you don't subordinate local naming contexts to the dnssrv, but rather
configure them as DNSSRV as well, and stack the chain overlay on top of
back-dnssrv; this way you have local naming contexts advertized and
directly accessed by direct searches, and __all__ DVSSRV, including local
ones, directly chased by the chain overlay.

p.

-- 
Pierangelo Masarati
mailto:pierangelo.masarati@sys-net.it


    SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497