Uniquely identifying a backend? (Was: commit: ldap/servers/slapd proto-slap.h)

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

Hallvard B Furuseth wrote:

> Howard Chu writes:
>  
>
> > Ando and I tripped over this before, the solution was to compare the
> > be->be_private pointer.
> >
> > I suppose it's possible that a backend (like back-null?) may have no
> > private data, so perhaps be_nsuffix is a better choice.
> >    
>
> Back-null has private data.  OTOH, back-passwd's be_private is only set
> if it is configured with an alternate passwd file, otherwise it's NULL.
>
> be->be_private need not be constant for the lifetime of a database:
> back-ldap/chain.c and overlays/translucent.c modify it.
>  
I have plenty of custom overlays that do modify it transiently.  That's 
one of the issues I was having when adding back-config capabilities to 
back-ldap & slapo-chain.

> back-monitor/monitor_back_db_open() reads be->be_private and stores
> it back, even though it never changes it.  I don't know why.
>  
It'll be cleaned up some time.  It was just, I guess, for consistency 
with other backends.

I think there might be cases in which all of the generic backend stuff 
gets changed for some reason, most likely during internal operations.  I 
wouldn't even trust be_suffix or be_rootdn (which might be NULL) or so.  
What about adding a be_tag field to the BackendDB structure that is 
unique (e.g. it gets some number from a given source when it's marked as 
a new database) so that it's sticky even when copying a backend 
structure into a temporary for internal operations?  To feed it, one 
could envisage a backendDB_tag(BackendDB *bd) function that is passed a 
pointer to a backendDB structure and univoquely marks it.  It would be 
called by backend_db_init() for each database allocated via the 
"database" config directive, but it could be called also directly to tag 
internal databases (e.g. in slapo-pcache, slapo-translucent and so).

p


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