To:
Eric Brunner-Williams in Portland Maine <brunner@nic-naa.net>
CC:
Rick Wesson <wessorh@ar.com>, "Hollenbeck, Scott" <shollenbeck@verisign.com>, "'Edward Lewis'" <edlewis@arin.net>, ietf-provreg@cafax.se, jaap@sidn.nl
From:
janusz sienkiewicz <janusz@libertyrms.info>
Date:
Mon, 17 Mar 2003 14:07:41 -0500
Sender:
owner-ietf-provreg@cafax.se
Subject:
Re: [ietf-provreg] thursday's meeting
Eric Brunner-Williams in Portland Maine wrote: > > I suspect because the granularity requetsted is for the registrant not the > > session, where by one session could do many registrations for many more > > contacts/registrants. > > So, in this model, there is p1, p2, p3; three distinct mumbles that have a > to be defined, but not session-scoped, definition. > > A session between e1 and e2, two EPP endpoints, which transports d1, d2, d3, > three distinct EPP datums, is: > > e1: session-open; send d1:p1; send d2:p2; send d3:p3; session-close > e2: ack, ack, ack, ack (sort of like the Martians in "Mars Attacks") > > If so, then while data exhibits a property close to global uniqness within > each glob of data, the value for the set {p1, p2, p3, ...} may have a much > smaller cardinality. Somewhere less than 42 seems likely. > > If so, then connections should resemble something like this: > > send d1:p1, send d2:p1, send d3:p1 .... send dN:p1 \ > send dN+1:p2 send dN+2:p1, send dN+3:p1 .... send dN+M:p1 > Shouldn't "send dN+2:p1, send dN+3:p1 .... send dN+M:p1" have p3 instead of p1? > > for non-trivial sequences of sequentially identical values of p. > > Rewritten, to show policy-scoped sequences of data: > > p1{d1, d2, ... dN}, p2{dN+1}, p1{dN+2, dN+3, ... dN+M} > > > And that looks reasonably amenable to this > > s1{p1{d1, d2, ... dN}}, s2{p2{dN+1}}, s3{p1{dN+2, dN+3, ... dN+M}} > > or three sessions, each with an associated single policy on all data. > > Operationally, we haven't precluded server operators from assigning well > known values of p to ports, so that in practice, the above would be three > distinct connections. > > We also haven't precluded server operators from cycling through a set of > well known values of p whenever the server observes a session-end sent > by a client, followed by a session-begin. > > We could of course add some mechanism for client and server to "dual" with > <dcp> values, until "trained", but until someone starts putting values for > p down, we don't actually know if the size of the p value-space is as big > as 42, or as small as two, or three, or four meaningfully distinct values. > > Eric > > P.S. I'm having way too much fun with the symbolic general nonsense. The > proponant(s) of mumble need to attempt to be concret, soon. Janusz Sienkiewicz