Re: The world's biggest data consumers don't think priority is such a big priority
RJ Atkinson <rja <at> extremenetworks.com>
2003-10-01 13:58:24 GMT
On Tuesday, Sep 30, 2003, at 19:46 America/Montreal, James M. Polk
> Needless to say, I completely disagree with your position and don't
> rely on quotes from people I don't know (like you do apparently) to
> get my information. I like first hand knowledge.
You claimed that no one was using Ethernet for critical services
without prioritisation (exact quote follows):
> The US Gov and others already employ priority networks; those that
> haven't moved their time critical applications over to their Ethernet
> networks (yet)
I have first-hand knowledge of several sites that do use Ethernet
for time-critical applications and do not employ prioritisation.
I work at such a site. I know of other such sites in various places
around the world. I'll take any data I can get. Different data
gets different weight in my analysis, depending on an assessment
of its quality.
Moreover, unless one is omniscient, which no one on this list
is, one could not know whether 'all of those that don't use priority
networks' do (or don't do) *anything*. There are far too many networks
in the world for anyone to know all the pertinent details about all
of them. Yet you claim to have just such knowledge.
Your claim was too extreme. Had you said "many" or "some" or
"few" rather than "those that don't", it might have been a lot more
plausible statement. However, you didn't do so and did not clarify
your meaning when questioned about it yesterday. Pity that.
> I was merely reacting to the quoted person's claim that prioritization
> in IP was not necessary.
Nonsense. There was no comment about "prioritisation in IP" in
that article. You badly misread the article at the URL provided.
(For any who tuned in late, the URL is:
1) The article at that URL does NOT anyplace contain the phrase "IP"
or the phrase "Internet Protocol".
2) The article more than once states that AT&T is discussing an
3) Ethernet is not IP, and is in fact at a lower layer in a
a conventional protocol architecture.
Now, as it happens, IEEE 802.1p (now packaged in a revised 802.1d)
specifies Ethernet precedence. In some ways, that IEEE mechanism is
similar to the IP precedence mechanism. I know of a number of sites
that use 802.1p and also a number of sites that don't use 802.1p.
Now, since a private email has erroneously suggested that my employer
would benefit from users not deploying QoS, I'll clarify that point.
My employer quite possibly has the *best* implementation of IP ToS,
DiffServ, and 802.1p available. Some of our customers enable some
of those features, some do not. It would be advantageous for my
employer if more folks enabled those features, because it would provide
additional crisp differentiation from many of our competitors.
However, on an IETF list, I express my own views about the standards
topics under discussion, rather than advocating standards positions
because they might advantage my employer. This might be confusing
to some. It does not confuse those who've watched me in the IETF
context over the years.
In any event, the issues with IP prioritisation on this IETF list have
focused on the inter-domain context, where deployment of inter-domain
DiffServ (or similar) mechanism creates a significant (Distributed)
Denial of Service attack vector not present otherwise. If anyone
has a technical approach that enables deployment of DiffServ in an
inter-domain context without creating a (D)DOS attack vector,
I'm sure it would be of general interest to this list (it probably
deserves a new Subject: line, however .
rja <at> extremenetworks.com