Unique NIC/RIPE/xxx-Handles

"Wilfried Woeber, UniVie/ACOnet" <woeber at cc.univie.ac.at> writes
 * | * Given that one of the aims of handles is (I think !) a more
 * | * compact, shorthand form of referring to someone, the  RIPE-DR222
 * | * idea sounds better to me - DR222 at rs.internic.net is longer than
 * | * my real name !
 * |
 * |I have to agree here. I think a nice syntax would be:
 * |
 * |<NIC>-XXXYYY
 * |
 * |where it is probably is too much hastle to ask InterNIC to redo all their
 * |nic-hdls to have INTERNIC-XXXYYY, so they probably keep the handles without
 * |the NIC. The others could be:
 * |
 * |RIPE-MT2
 * |APNIC-MT2
 * |...
 * 
 *  My original reasoning was something like:
 * 
 *  Keep the InterNIC-Handles as they are, ie. when there is no NIC-ID, then
 *  let's assume it is InterNIC. For the others, tag them. For the DataBase,
 *  store everything without a (external) tag with an (internal) InterNIC tag.
 * 
 *  Also, I think it is really not necessary - maybe not even useful? - for
 *  people to obtain more than one handle. However if someone is hit by that, o
 * r
 *  even wants it :-), we could probably live (DataBase-wise) with a definition
 *  of (REQUIRED, MULTIPLE, UNIQUE).
 *  				      (Marten: could the DB-SW digest this?)

The thing that is tricky here is that the combination of name and nic handle
is used as the unique key for a person, and these fields cannot be multiple
because then I cannot guarentee the ordering of the unique key any longer.
What I mean is say I have two handles:

person:  Marten Terpstra
nic-hdl: MT2
nic-hdl: RIPE-MT2

The unique key could then be:

"Marten Terpstra\tMT2\tRIPE-MT2" and
"Marten Terpstra\tRIPE-MT2\tMT2"

which are different unique keys for the same person, which gets me into
trouble. On second thought, with the new software:

person:  Marten Terpstra
address: RIPE NCC
address: Kruislaan

and

person:  Marten Terpstra
address: Kruislaan
address: RIPE NCC

are different objects, and if the unique keys are the same, this will cause an
update. (Am I still making sense ?) Inside an attribute the ordering does
matter .... (These are the sort of things that come up implementing all this
stuff)

The more I think about it, the more I like the idea I said in my last mail,
ie same global nic handle space, just divide the numbers that go with the
initials ...

-Marten