linked attributes, DRS and abusing the ldb modules API

[Andrew Bartlett]

On Wed, 2009-08-26 at 22:18 +1000, tridge at samba.org wrote:
> Hi Andrew,
> 
>  > >  2) we could call into the linked_attributes module from the
>  > >     repl_meta_data module, letting the linked_attributes module do its
>  > >     work as usual when it gets a modify request on an attribute that
>  > >     is linked
>  > 
>  > The problem with this is that it fails across module boundaries.  The
>  > repl_meta_data module operates after the partitions module, so I'm
>  > worried what happens to a link that crosses the partitions.
> 
> I don't think that is a problem. When the repl_meta_data module is
> called, its module chain looks like this:
> 
>   repl_meta_data->tdb
> 
> the ldb_load_modules_list() call I'm making changes it to this:
> 
>   repl_meta_data->linked_attributes->tdb
> 
> so the partition module is not being linked in. We're just adding the
> linked_attributes module to the list of those modules that should
> process the modify request we've constructed.

Sure - but when you link to the other partition, how will the
search/modify work to create the backlink?

>  > It's an interesting hack.  We also have to get in some of the other
>  > modules, like the extended DN storage stuff.   
> 
> I'm not at all sure what modules we want to get in there. The whole
> design of the DRS pull replication code seems to be based on bypassing
> modules. It could just call a ldb_modify() at the top level, but
> instead it constructs a complex extended operation specially designed
> to bypass all the top level modules and jump down below the partition
> module. I'm guessing that Metze did that for some reason (as Metze
> understands this DRS stuff a lot better than I do!).
> 
> The problem is that it does need the linked_attributes module, and the
> API usage I've proposed is the only sane way I can see to allow a
> caller to call into a specific list of modules.
> 
> If it turns out we also need the extended DN module, then we can add
> it to the list. That's why I made replmd_ctx_init() take a char **.

OK.  We actually get the extra metadata (the extended DN parts) over the
wire, so perhaps we don't need them.

Andrew Bartlett

-- 
Andrew Bartlett
http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org
Samba Developer, Cisco Inc.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20090826/8e5f7eca/attachment.pgp>