[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Re: Submission of Internet Draft - Complex Directory Lookup using Javabased LDAP Query Extension
At 17:43 07/01/99 -0600, you wrote:
>Perhaps rather than sending code to a server it would be reasonable to
>identify an extension to the filter syntax so that nested searches could be
>expressed in a manner similar to the nesting of 'selects' within an SQL
>'where' clause.
I think this is the right general approach. I disagree with the breadth of
assertion in <draft-goenka-ldapext-java-complex-search-00.txt>:
3.2 Proposal Overview
The key reason for the above problem is the inability to run a
piece of client code securely at the server end. [...]
I think the 'key reason' here is the limited expressive power of the search
filter, as currently defined. I think that to take on the full complexity
of a generalized programming environment for this purpose is not required;
I like Java, but in this instance I think its a hammer looking for a nail.
I would suggest that the key area to examine would be the right hand side
of a search filter primitive. From RFC 2254:
item = simple / present / substring / extensible
simple = attr filtertype value
filtertype = equal / approx / greater / less
equal = "="
approx = "~="
greater = ">="
less = "<="
:
attr = AttributeDescription from Section 4.1.5 of [1]
matchingrule = MatchingRuleId from Section 4.1.9 of [1]
value = AttributeValue from Section 4.1.6 of [1]
In my view, if the allowable 'value' can be generalized in some way to
reference a result obtained by some other lookup, a whole range of complex
directory lookups could be accommodated without requiring a full-scale
programming invironment in the server.
For those servers that do happen to have JVM environments, I think that the
extensions could be designed to be easily mappable to a Java subset so that
the JVM becomes a resonable (but not required) implementation technique.
But, whether this is appropriate functionality to consider for an LDAP
server is a queston that needs to be considered first, before getting too
hung up on design details...
#g
------------
Graham Klyne
(GK@ACM.ORG)