[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


To: janusz sienkiewicz <janusz@libertyrms.info>
cc: Eric Brunner-Williams in Portland Maine <brunner@nic-naa.net>, Rick Wesson <wessorh@ar.com>, "Hollenbeck, Scott" <shollenbeck@verisign.com>, "'Edward Lewis'" <edlewis@arin.net>, ietf-provreg@cafax.se, jaap@sidn.nl, brunner@nic-naa.net
From: Eric Brunner-Williams in Portland Maine <brunner@nic-naa.net>
Date: Mon, 17 Mar 2003 14:16:13 -0500
In-Reply-To: Your message of "Mon, 17 Mar 2003 14:07:41 EST." <3E761CFD.9E18F2F@libertyrms.info>
Sender: owner-ietf-provreg@cafax.se
Subject: Re: [ietf-provreg] thursday's meeting

> Shouldn't  "send dN+2:p1, send dN+3:p1 .... send dN+M:p1" have p3 instead of
> p1?

I should have been explicit.

I'm drawing on some non-provreg data -- real life with cookies.

Few distinct "privacy" "policies" or "preferences" exist.

The distribution of policied cookies and browser values that result in
these three general categories:
	accept-unconditionally
	warn-accept-or-reject-conditionally
	reject-unconditionally
is small, and the modality is less than the total range of values.

The most likely transition will be from one domainant mode to another.

So, p1 to p2 and back to p1, generally, and p1 to p3 and back to p1.
Less frequently will be pN to pM, where N and M are not in the very
small set of dominant modalities of meaningfully distinct policies.

I wasn't trying to show p1, then p2, then p3 usage, but
	
 ...p1p1p1p1p1<break>p2p2<break>p1p1p1p1p1p1pp1<break>p3<break>p1p1...

Where there really is random variation from one data glob to the next, and
the variations are meaningfully distinct (a kettle of fish that is not yet
opened), then clearly the "statefull encoding" of policied sequences of
data globs is not going to work as well as some "statefull encoding" of
policied, and uncorrolated data globs. 

In a sense, we're just revisiting statefull vs stateless encoding design
issues. How much character-set variation do you think there is in some
character stream? Design accordingly. If few transitions relative to the
byte count, go for what amounts to state shifts (and variable byte counts
per character). If many relative to the byte count, consider using fixed
rather than variable byte counts, and accept the "padding" overhead.

How many interesting ways are there to write: don't publish some or all of
my mumble?

Eric

Home | Date list | Subject list