[Date Prev][Date Next]
RE: Plea for Server Side Sorting
Unnecessary to one, necessary to another. When we load 1 million concept codes from large terminologies into an ldap server, sorting them on the client side is not feasible. The sorting should be done on the server - because the server can do things the client can't - it can pre-compute the sort order and it typically has more memory/processing power for doing the sorting.
I know that IBM's Tivoli supports sorted results on large data sets that behave as you expect they would. Of course, that's not OpenSource... Just making the point that sorting results isn't a limitation of LDAP design in general - its more a limitation of the underlying implementation of OpenLDAP. I imagine if it were easy to implement, you probably wouldn't get such a strong pushback from your offer.
From: owner-openldap-software@OpenLDAP.org [mailto:owner-openldap-software@OpenLDAP.org] On Behalf Of Kurt D. Zeilenga
Sent: Thursday, March 31, 2005 11:58 PM
To: Eric Nichols
Cc: Owen DeLong; Dave Horsfall; OpenLDAP Software List
Subject: Re: Plea for Server Side Sorting
At 07:24 PM 3/31/2005, Eric Nichols wrote:
>I'm just trying to bring something positive to the table here. I wouldn't ask
>for it unless there was something like an RFC and client ability to use
>sorting to help point the way to a solution.
I suggest you read the RFC again. Note that the RFC doesn't
require the server to return sorted results. It only
a method for a client to request sorted results and a server
to indicate that the results are sorted or not.
>In my limited logic and knowledge of OpenLDAP/slapd internals...
>No matter how complex the query (using indexing or not) there is still
>produced a result set. Maybe someone can enlighten me, but I'm guessing that
>result set is stored in memory before it is sent to the client (or paged to
>I do not have any doubt in an ldap application wanting to make an API call to
>a server... If I am over simplifying please tell me how it can be done, not
>how it can't...
Robust directory clients that need sorted results sort the
results themselves. Robust directory clients avoid extensions,
especially extensions that add unnecessary complexity.