[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: >= (greater or equal) and <= (lower or equal) operators in

Hallvard B Furuseth wrote:

Possibly, but more likely they just mimiced how how most other
attributes are defined: Almost none of them has ORDERING rules, even
common ones like 'name' and thus 'cn' - and noone can possibly know all
the purposes those two will be used for. I've asked for ORDERING rules
for some attributes once in a while too, but usually too late.


Really, I have no good reason to defend that schema; what I'm saying is that
I think it was designed with some objective in mind, and that didn't require
the ordering match. Of coursethe same attribute could be used for other
purposes and required the ordering match, but then another attribute might be
needed. Someone is doing something wrong: either who first designed it was
short-sighted, or who's trying to reuse it is too ambitious. I wouldn't blame
the short-sighted, then, because it always easy, AFTER, to say that someone
was short-sighted.

Is it considered breaking standard schema to add ORDERING rules into
them?


I would love to and I'm about to throw in the towel and "break" this
standard schema by adding this (and other) rules. A broken or
difficult standard is of no use.


Then just design your own, or propose a draft for standard emendation,
but don't hijack others, or you'll run into interoperability problems,
sooner or later.



True, but sometimes the attributes are needed for another application
which expects a particular attribute name like uidNumber (e.g. PAM, I
think).  Then you'd have to store that other attribute in _addition_
to uidNumber.

Or you could make your own schema with the same attribute names, object
class names etc. as the Pam schema, but with your own OIDs and with the
matching rules you want. That's 'formally' OK - unless the RFC2307
attribute names are registered with IANA - and a bit cleaner than
modifying the original schema, but still rather ugly, I guess.


Or one could design a new attribute that inherits from the old one PLUS the
ordering match; inheritance wwould let the new attribute be a valid replacement
for the old one without even hanging a line of code (except where writes occur,
of course) and add the new feature (an mess everything up, of course).

Ciao, p.





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