[1248] in Commercialization & Privatization of the Internet
Re: technical details
daemon@ATHENA.MIT.EDU (Martin Schoffstall)
Wed Aug 28 13:09:22 1991
In-Reply-To: <CMM.0.90.2.683328595.ittai@shemesh.ans.net>
Date: Wed, 28 Aug 91 11:35:53 -0400
To: "Ittai Hershman" <ittai@shemesh.ans.net>
Cc: "William Schrader" <wls@psi.com>, com-priv@uu.psi.com, lear@turbo.bio.net
From: "Martin Schoffstall" <schoff@mail.psi.net>
Reply-To: schoff@psi.com
I don't know where to begin.....
>DATE: Tue, 27 Aug 91 17:29:55 EDT
>FROM: Ittai Hershman <ittai@shemesh.ans.net>
>
>Similiarly, ANS is studying the CIX agreement in detail. Some of
>our concerns with regard to ANS joining the CIX, are:
>
> o The CIX may not be scalable because it does not provide
> routing arbitration.
What does routing arbitration mean to you? If we don't do it, then
certainly you don't either. Now MERIT does do routing arbitration for
the global Internet under contract to the NSF, now hopefully this will
be moved to the neutral NEX and be done by similiarly qualified people
for the entire Internet.
Now from my perspective the CIX does do a manner of routing arbitration,
but more importantly all members do also. In fact all regionals with
connections to "other networks" do routing arbitration.
Now let's go back to what MERIT does for a moment. Due to the arbitration
methodology that they chose, in 1987 we lost dynamic routing in the "core".
Now we get to wait 3-4 days for routing updates to be done (for US sites),
for international sites it is MUCH worse. Up until then we had dynamic
routing - we lost something in this grand transition.
For the future we need to do better; hopefully we will in the NEX, I
think a mixed model of dynamic and static routing needs to be seriously
considered.
> The CIX is merely an interconnection
> point among a set of autonomous systems.
Maybe. But it is and will be evolving based on the needs of the membership.
> To some this
> is a feature, to others it is a serious design flaw.
> Is the CIX model as scalable as its proponents suggest?
We believe so. But given our experiences of many years we tend to lead
and create what we need. So when we need a quarter inch
widget we go down to the workshop and whip one up, or a couple thousand,
or contract out for a couple million.
>
> o The flat-fee settlements (previously described as a no
> settlements policy) only works among equals.
>
> The CIX founders are more-or-less equals at least in terms
> of network bandwidth. Some have wider-scale deployment than
> others, but the differences among them do not include the
> infrastructure bandwidth. ANS is different in this regard.
> We have a coast-to-coast T3 backbone.
You assume that the metric for determining the cost of a network is how
many bits/sec. Quite frankly for those of us in the retail business,
the people cost are what is expensive - to do the training, support, and
operations.
Just because you went out on speculation and built a big network that
costs you a lot of money (actually I believe NSF pays how many millions
of dollars?) doesn't apply at all. Normal commercial settlement (yours
is far from normal) in all other telecom areas is either based on actual
usage or is flat fee. Just because you have a big pipe doesn't mean its
being used.
But let's work with your belief system, its still moot because right now
we're all working with T1 Pipes into the CIXNet, so we all arrive at
equal footing. now the funny aspect here is that the last word I heard
was that you were talking about a 56kbps connection to the CIX, at least
that is what was said at the FARNet meeting. So even in this your not
real consistent.
>
> If we were to interconnect with the CIX, how could we ensure
> that other CIX members didn't piggyback off our infrastructure?
>
Why would we want to? To do what?
> Can the CIX guarantee that only legitimate inter-carrier
> traffic (i.e. CIX affiliated network customer to ANS customer)
> crossed the CIX to ANS interconnect? Using what methods?
Contracts, routing announcements, and access lists (you call them
"policy based routings").
>
> Perhaps the CIX should join ANSnet?
>
> o If commercial traffic would be sent across the CIX into ANS
> with no settlements, how would ANS fund the national and
> regional infrastructure pools mentioned in the "Mid-level's
> Guide"?
This is truelly noble, but from where did this mandate come from? Isn't
infrasctucture being funded through other means right now? Will it stop
if you dissapear? Quite frankly I think the mid-level infrastructure is
being funded by their subscriptions, the CIX and national network
providers are funding a national infrastructure. NSF is funding a
national infrastucture (with all the standard cavaets).
>
> ANS is committed to the concept of helping defray the costs
> of RE networking, by providing commercial attachments.
>
So is everyone. There are a myriad of methods how people are doing
this, your rhetoric is such that your the only one.
Marty
