Carol, Quoting from Carol Orange's message:
We actually discussed this point a bit in a slightly different light. An important question is: should the forward type be up to the object maintainer, to the whois server providing the referral (in this case RIPE), or the whois client, or end user.
This is a common problem in any distributed application (ie: DNS, X500, ...) Any of the party involved may wish a certain behaviour by other parties but the actual behaviour of the system is based on matching wishes with options chosen by implementors/administrators. The matching is usually done via a protocol. Proposing that the behavior of one of the components, the server, could be decided by the object maintainer is just a way to give a certain degree of autonomy to him.
As proposed here, it is the whois server, but could be modified in the future to also be the whois client.
The object maintainer can in fact always prevent automatic request forwarding by putting the referral information in a remarks field.
Yes, if you reduce to two cases only: referral and [chaining,forwarding] However it has to be said that the referral method is more robust and can scale: the server could be a bottleneck/point of failure
However, if such a mechanism should become popular, then it may be suitable that whois clients be developed that can parse the refer field and resend the request to the appropriate server. If we allow the object maintainer to determine this in the <forward-type> field, then it actually limits flexibility in the future.
No, if the server getting requests from "powered clients" decide that the wishes of the client take precedence over those of the object maintainer
Or is my thinking twisted?
-- Carol
---------- ---------- Antonio-Blasco Bonito E-Mail: A.Bonito@cnuce.cnr.it Reparto Applicazioni Telematiche c=it;a=garr;p=cnr;o=cnuce;s=A.Bonito CNUCE - Istituto del CNR Tel: +39 (50) 593246 Via S. Maria, 36 Telex: 500371 CNUCE I I-56126 PISA Italy Fax: +39 (50) 904052 ---------- ----------