On Tue, 9 Oct 2001, Wilfried Woeber, UniVie/ACOnet wrote:
Dear George,
I am worried that there may be *external* referencees to objects, and while we have no 'contractual' obligation it might not be nice to do this to people without some engagement but I expect most people would welcome losing a path to spam.
without going into any detail of the proposal itself, I perceive the fact that there are external references (allowed) _without "consent" and without documentation_ a major flaw in the architecture.
Unfortunately (and unlike HTTP), there is no neat 'referer' field passed by a WHOIS client to the WHOIS server that would allow any WHOIS Server operator to identify the objects in remote databases with references to local old/outdated/expired objects. Then again, it is not the problem of a given WHOIS Server operator to maintain referential integrity with databases outside their control unless there is an agreement specifying that there will be referential integrity. (ie, I don't think that it is an issue for most installations of the NCC database server software ;) )
As to automatic sweeping, I think there are problems with link counting which are similar to those we've faced in discussing cross-database linkage and references. Too many risks of the link count getting out of whack.
Fair enough, a very valid concern! But I think the proposal is trying to describe the "user view" and the "abstract behaviour".
Actually, the RIPE database software (both v2/RIPE181 and v3/RPSL) already have existing mechanisms to detect the existence of links (ie, cannot 'delete' an object while it is still being referenced by #X objects of type 'Foo' etc ) to an object by objects within the same database. So, keeping the link count up to date and correct is a simple (but non-trivial for the coders) task that can be attached to the existing create/modify/delete operations. (ie, increment/decrement the link count on all objects currently or no longer referenced by this object we're currently creating/modifying/deleting).
I think you could use it as a guide to a sweep, but not as a determinant without some review and other checks. Eg to trigger more exhaustive checks to see if there are are references in twisty ways.
I agree.
A previous iteration of this issue in the db-wg can be seen at: http://www.ripe.net/ripe/mail-archives/db-wg/19990401-19990701/msg00003.html http://www.ripe.net/ripe/mail-archives/db-wg/19990401-19990701/msg00007.html http://www.ripe.net/ripe/mail-archives/db-wg/19990401-19990701/msg00008.html -- Bruce Campbell RIPE Opinions expressed in this mail are my NCC own and are not that of my employer's. Operations