ietf-smtp
[Top] [All Lists]

Re: SMTP QoS

2001-07-08 16:08:20

Continuing to try to focus on the core concern:

At 12:54 PM 7/7/2001, ned(_dot_)freed(_at_)mrochek(_dot_)com wrote:

>>However, note that we still do not have a way to specify QoS at the lower
>>layers and that all efforts to provide that capability over the last 30
>>years have failed.
>
>And what makes you think that exactly the same thing won't happen here if we
>turn a simple priority label into a full-featured QoS scheme with tightly
>specified semantics?

I think you flipped the high-order bit on the semantics of my comment.

The 30 years has been marked by a syntactic field that has had no semantics
standard.  That is what you are supporting now, though the 30 years says it
won't be productive.

This isn't my area, but the lower layer folks I've heard talk about QoS
problems don't appear to agree with this analysis or conclusion.

IPv6 is attempting to provide detailed semantics.  Last I heard, that is
both difficult and believed to be essential for v6 QoS.

Again, this isn't what I've been hearing. Only today I heard from one of the
IPv6 folks saying that what's needed is a simple, lightweight mechanism that is
actually implementable and doesn't promise more than it can deliver, and that
this is what has been missing all these years.

> I believe that a less nailed down mechanism is potentially quite useful as a
> tool that MTA implementors and service providers can use to build various
> services. It is NOT a service in and of itself,

This is an interesting approach to Internet standards.  Let's define field
syntax, but defer defining standard semantics until the industry
experiments enough to converge on a specific.

Nowhere did I say that there need be no absolutely semantics associated with
priority. Such a notion is clearly absurd, since it would allow someone to
assign opposite meaning to priority labels, or to assign different meaning
depending on the day of the week, or whatever.

The rest of your "argument" is nothing more than a straw man derived from this
gross exaggeration, so I'm not going to bother to refute it. And as for
examples of standardized things with incompletely specified semantics, there
are so many examples of these in successful standards I don't know where to
begin.

Apparently we ARE ready for private efforts at exploring preferred
semantics.  That's excellent.

And in order to conduct such efforts you have to have a way of labelling
things. That, IMO, is what Julian is trying to create.

                                Ned


<Prev in Thread] Current Thread [Next in Thread>