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

Re: ldapbis WG Last Call on ldapbis-syntaxes, ldapbis-strprep

Hallvard,

The way you worded your suggestion, the suggestion applies to all
currently prohibited characters.  That is problematic as prohibiting
certain assigned code points as well as all unassigned code points is
quite appropriate.  (Note that as code points are assigned, one can
update the profile to properly handle them.)

If you limit your suggestion to just private use code points, then
I think we might be able to change the algorithm such that those
points may be handled in a "local manner" (meaning that the
implementation may map, normalize, prohibit, bidi check,
insignificant character remove as they please) but we'd also
need to not only state some interoperability and security
considerations.

I don't necessarily support making such a change.  I offer it for
discussion purposes.

Kurt


At 12:16 PM 6/27/2003, Hallvard B Furuseth wrote:
>Kurt D. Zeilenga writes:
>>> In view of this, I'm beginning to dislike that Prohibit step.
>>> If I have a purely local LDAP directory, why shouldn't I be allowed to
>>> give an attribute two values where one uses a Private Use code point?
>> 
>> These rules are designed for interoperability between independently
>> developed implementations which you are likely to have even in
>> local environments.  (...)
>
>We have at least two LDAP servers here (OpenLDAP and AD), but will never
>need any interoperability between them.
>
>Interoperability is important, but I think the price is too high
>sometimes.  Like in this case.  I don't like to have to pay too much for
>features I know I won't need.  That's why I think the prohibit step
>should only be a SHOULD or RECOMMENDED, with a 'MAY allow admins to turn
>it off'.
>
>> Note that one is always free to define other schema elements to meet
>> local needs.
>
>That only helps if I can hack the server to add new matching rules, or
>if I'm satisfied with using Octet String for string data and don't need
>features like case-ignore matching.
>
>-- 
>Hallvard