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


To: Patrik Fältström <paf@cisco.com>
cc: "'ietf-provreg@cafax.se'" <ietf-provreg@cafax.se>
From: Eric Brunner-Williams in Portland Maine <brunner@nic-naa.net>
Date: Fri, 27 Sep 2002 11:50:58 -0400
Content-ID: <14627.1033141858.1@nic-naa.net>
Sender: owner-ietf-provreg@cafax.se
Subject: Some musings.

Morning-Afternoon Patrick,

Taking the "... acts like UDP ..." and "... issues with SMTP..." as our
starting points for this morning's koffee-klatch. Please pull up 1122
in your editor of choice. I assume you are wearing your AD hat.

A transport does bookkeeping and retransmission, and eventually offers
...  8-bit binary stream semantics ... fine.

But what part of this resolves the issues?

Now move down to page 108 in 1122 and mull over the requirement summary.
One or more of these is missing from, or denoted differently in, at page
80, for UDP.

How far does [lost,retransmission,ordering,etc.] of application-layer
program data chunks extend?

What does [secure] mean?

I can't read your mind.

Suppose we have SMTP over TCP. Presumably whatever our concerns are, they
are not in the solved-by-stream-semantics problem domain. Something along
the lines of end-to-end application-layer semantics, like session state.

If we have SMTP over UUCP (and why not?) 1122 falls out of the picture,
or at least its temporal guarantees within the boundaries of a socket's
connection life-time. We still get end-to-end application-layer semantics,
with store-and-forward best-effort replacing the (gatewayed best-effort) 
single socket.

So, idiot that I am, I've talked (typed actually) my way into thinking
that the issues that you could be thinking of would be solved if the EPP
over FOO transport binding (optionally) provides for the following:

	o a mechanism to provide a UID (or non-wrapped in reasonable time),
	o a mechanism to provide for non-repudiation of origin,
and
	o a mechanism to provide for non-repudiation of receipt.

I think that covers [lost,retransmission,ordering]. 

I think "secure" could mean providing for the following:

	o encrypting data (optionally)
	o signing data (optionally)

Naturally, the EPP over FOO transport binding might have something useful
to say about formats used too.

Let me know what I've missed. Back-of-the-moon kind of stuff. TiA.

Cheers,
Eric

P.S. I'm fond of 1009, 1122 and 1123. I hope you'll forgive my preference.
     I'm also fond of UUCP. 3335 is worth a moment's notice also.

Home | Date list | Subject list