directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Howard Chu <>
Subject Re: [OT]Einstein was right ...
Date Thu, 29 May 2008 01:12:19 GMT
Graham Leggett wrote:
> Emmanuel Lecharny wrote:
>> .. when he said that two things were infinite :
>> - the Universe,
>> - human stupidity
>> But he added that he doubts that the first assertion was right.
>> Regardless, the second one seems to be correct :
>> How possibly some idiots can patent such an incredibly piece of garbage ???
> Prior art kills patents.
> Both caching and reverse proxying have well documented prior art in
> httpd going to before 1999 (when I first got involved with it). It would
> have to be determined whether HTTP and LDAP differ enough to argue that
> the work is original. (Both HTTP and LDAP are request response protocols
> where an end user makes a request for information and receives a response.)
> Alternatively, I am sure that Sun produced an LDAP proxy server capable
> of doing directory aggregation about a decade ago, may be good to check.
> Anyone remember?

Sun didn't have one back then. I wrote back-ldap for OpenLDAP in 1998 though; 
that was my first contribution to OpenLDAP in fact. The rest of the world 
wasn't even thinking about LDAP proxy servers at the time, they were too busy 
building LDAP islands that didn't talk to each other.

I would say that the technique used to perform LDAP query containment was 
interesting and certainly novel, wrt to HTTP. But the problem with this patent 
is that it was granted in 2006, and we had already implemented this in 
OpenLDAP in 2002.

Apurva Kumar, who developed the first version of the OpenLDAP code, is also 
the inventor on both of the above patents. Since he's an IBM employee, it's 
not surprising that he would have applied for patents (in general), since 
that's common practice at IBM. But someone ought to have told him that once an 
invention has been shared with the world, it is no longer patentable.

I wouldn't call query containment "garbage" - LDAP search queries can be 
pretty complex, and it certainly takes some thinking to get caching right. 
Give credit where it's due. I don't believe in software patents though; this 
is all pretty stupid. And of course, Apurva's original implementation was 
pretty atrocious; I had to rewrite about 95% of it before it would offer any 
performance benefit over no cache at all.

The other thing about patents is you used to have to submit an implementation 
along with the app; i.e., you can't just concoct an idea and send it in. You 
have to demonstrate that the idea is practical in the real world. And, you're 
supposed to submit the best possible embodiment of the idea. If he used his 
original OpenLDAP contribution in the patent app, it's clear that it was 
nowhere near the best possible embodiment. Oh well.

The second patent is essentially one step beyond syncrepl and other similar 
replication mechanisms that have existed for many years. We've had partial and 
fractional replication for ages. The difference that I see here is that he's 
using query containment to decide to return a referral to the master server, 
when a searched-for entry isn't present in the replica. In our current 
implementations, if you ran such a query against a partial replica, you'd just 
get no result back for the missing entries. As patents go, it follows the 
standard formula - add one step to an existing well known process and presto, 
you've invented something new. Whether it's actually useful or not is a 
different question - IMO, anything that relies on LDAP referrals is 
fundamentally flawed anyway.
   -- Howard Chu
   CTO, Symas Corp. 
   Director, Highland Sun
   Chief Architect, OpenLDAP

View raw message