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

Re: Final pieces to the Netscape roaming puzzle found!

At 11:13 AM 7/25/99 -0400, Kartik Subbarao wrote:
>Agreed. I've gone back through the OpenLDAP list archives and reread
>some of the discussion. I have posted a message to some Netscape and
>Mozilla newsgroups raising the awareness of the two major roaming
>issues, and pointing out how Netscape could interoperate better with
>OpenLDAP. (Mozilla doesn't seem to have any roaming code yet).

I think we need to follow the old rule: be strict in what you
generate, be liberal in what you accept.  This specific
issue highlights an example where the OpenLDAP server could be
more liberal and the Netscape client more strict.

However, here are some LDAPv2 vs. LDAPv3 differences to which this
rule cannot be applied.  It particular, LDAPv2 uses T.61 (or ASCII)
encoding of strings where LDAPv3 uses UTF-8.  The implementation
either uses the required charset for the protocol or doesn't.  When
you mix an implementations which do abide with implementations
that do not, significant interoperability problems will arise.
This issue will likely cause the LDAP community a great deal
of heartburn.

Though not yet implemented in -devel, we intend to implement
to charset translation such that strings are encoded properly
for each protocol.  That is, we'll store UTF-8 and convert
to/from T.61 for LDAPv2.

>As you've seen here and as I mention in my newsgroup posts, I don't know
>too much about LDAP precedent. Could you explain why generating the
>"modifyTimestamp" attribute upon creation might not be a Good Thing in
>the LDAP world?

What's important is that each attribute type has well defined syntax
and semantics.  In this case, the semantics of modifyTimestamp are
defined by X.501(93).  Implementations which provide different
semantics may not interoperate with implementations that expect
semantics as defined in the specifications.

>I'm coming at this from the filesystem school of
>thought, where creation and modification timestamps are both initialized
>when files are created.

You need to check your luggage before boarding...