More Sidgemore on per-bit pricing (fwd)
paul at vix.com
Tue Dec 8 04:20:19 UTC 1998
> > One could use something like MPLS to determine the Ingress and Egress
> > points of the network if it had a robust enough MIB.
> Goodbye scalability.
> Please tell me what happens when a link carrying a million MPLS-ed
> flows does flap.
I can't. But scalability isn't important in the modern era. MPLS and to
some extent QOS are all intranet standards, and if they are ever applied to
the true universal Internet in an end to end way we'll see a pretty much
vertical rise in the complexity of things like BGP and record-route and
traceroute. No matter how bad an idea it seems like, the fact remains that
to push public data networking to the next level of functionality, vendors
and pundits and especially venture capitalists are deliberately waffling on
the definition and extent of the word "public."
Stateful internets are bad. Stateful intranets aren't so bad. The last
voice-over-IP story I heard was just a way to increase cost efficiency in
the last mile -- it wasn't an end-to-end thing. The one before that was
just to increase cost efficiency in a long distance carrier -- it wasn't
an end-to-end thing.
Vendors of both equipment and services are thinking of new technology in
terms of what it can do for large VPN-consumers or for networks which mostly
just move data for their own direct customers. And of course, most of the
so-called first tier expects there to just be one network in the endgame,
with all others as customers of it.
The people who, six years ago, yelled about NAT, and ten years ago, yelled
about firewalls, all because "internet" means "network of networks" and
anything which couldn't scale to universal end-to-enditude shouldn't be
used (and certainly not standardized!) probably had the unspoken worry
that we were headed for a "network of VPNs" or a "network of NAT clouds"
or a "network of its own customers".
I had to think my way through all of this in order to understand the last
batch of Internet industry investment analysis briefs I slogged through.
Nobody, anywhere, expects to see a million MPLS'd flows going over a single
link before 2004, and by that time, the problems will be very different from
Paul Vixie <paul at vix.com>
More information about the NANOG