Re: Aliasing?

peter (
Mon, 15 Jul 1996 15:29:02 -0400

On Mon, 15 Jul 1996 wrote:
> > Mark Smith sez...
> > What Eric says is true -- the LDAP protocol, currently at version 2,
> > does support aliases. In the U-M LDAP 3.3 release, ldapd supports
> > aliases but the standalone server (slapd) does not. I can't think of
> > any easy workaround (aside from writing some code to enhance slapd).
> AH HA. sigh. I've been playing with building very nearly the same sort of DIT
> structure that Eric mentioned... on a slapd-based directory. Had been wondering
> if I'd done something wrong with my alias entries.
> Hmm. Ok, I'll bite, what are the plans for future public-domain reference
> implementations of LDAP and do they include Alias handling? "How much" work is
> it to impl that support -- like, roughly which of slapd's *.c files would have
> to be touched?

If this is done, I hope it is done "right", in the sense that aliases
should simply be alternate names for the same entry, and should be
transparent to the user. One of the mistakes (IMHO, a mistake, others
might call it something else) in X.500(1988) was making aliases "special",
in the sense that one could choose to treat them differently from
"ordinary" entries, and that they could not be used in all operations
(i.e., updates required the real DN). X.500(1993) corrects this by making
an entry's aliases more-or-less equivalent to the entry's "real" DN (i.e.,
one can perform a modify operation using an alias); unfortunately, some
of the "mistake" baggage had to be retained.

Simply put, aliases are misnamed: they should simply be names. Entries
start off with (at least) one, and more can be added later. Entries can
be accessed via any of the names, none are treated preferentially. The
best analog I can think of is that of hardlinks under the various UNIX
file systems.