RE: Search over referrals (Re: LAST CALL: draft-ietf-ldapext-ref erral-00.txt)

[Alan Lloyd <Alan.Lloyd@OpenDirectory.com.au>]

Thanks for the question.
We do not distribute indices/indexes - no need.
The DXserver design is subject to patents re the rdb design. Its a
question of using primary keys and logical indexes for the X.500 stuff
plus some other hush hush things. Its very obvious that the theoretical
X.500 hierarchial object model must be extracted into scaleable
searcheable(new word) relational parts - otherwise distributed
performance will be slow. ie we can do "not founds" in our DIB instantly
and pass this on via DSP if approprate - eg find three mobile phone
numbers from the top where they are all held in an adjacent DSA.
ie the better the database performance and that includes access control
checking the quicker the DSA distribution can be.

Indices and their distribution seem to be yet more engineering and
operational overhead when good RDB design does it all (under the
covers).

In addition global directories are about taking business information
which is traditionally held in RDBs and providing them with access,
distribution and replication under a globally object oriented paradigm
(ie,. X.500) Its just that most of the directory work to date has
dropped the storage integrity issues gained by RDBs and put the
directory info into filespace ie. Lost all the effort by the global
vendors of RDBs - which are essential for large scale, high integrity
business (distributed) information systems.


ie. I view directories as communications enabled object oriented
databases, not communications protocols for information access. ie
protocols to the minimum, performance and integrity to the max.


Off list discussions are OK re our technology and its design.

regards alan
> -----Original Message-----
> From:	dboreham@netscape.com [SMTP:dboreham@netscape.com]
> Sent:	Monday, April 06, 1998 2:19 PM
> To:	Alan Lloyd
> Cc:	ietf-ldapext@netscape.com
> Subject:	Re: Search over referrals (Re: LAST CALL:
> draft-ietf-ldapext-ref	erral-00.txt)
> 
> Alan Lloyd wrote:
> 
> >         We thought about indexing in fact we use an RDB and our
> table
> > design to do that for every attribute - automatically. We also have
> 
> I believe Tim was talking about distributing indices
> between servres, rather than re-indexing on each server.
> If you rely on a commercial relational database
> to distribute indices, then presumably it's using
> some proprietary protocol to do so  ?
> 
> Sounds interesting anyway, I'm fascinated to hear
> how it works. Do tell.
> 
>