North American Network Operators Group|
Date Prev | Date Next |
Date Index |
Thread Index |
Author Index |
RE: CIDR Report
- From: Roeland M.J. Meyer
- Date: Sat May 13 17:52:04 2000
> Danny McPherson: Saturday, May 13, 2000 1:47 PM
> > None of these are big enough to justify their own backbone
> operations or to
> > buy a backbone from someone else, or there wouldn't be a
> problem. Paying scads
> > of extortion money is also problematic (cheaper to simply
> burn the IP addresses).
> > I am NOT advocating tossing all of that out. I am simply
> bringing up a
> > problem condition. Please, don't shoot the messenger, or
> otherwise get
> > defensive (return fire is a bitch).
> Nope, all of these are reasonable, the ones that aren't are,
> for example,
> where folks have a single connection, or multi-home only to a
> single provider.
Agreed, peering on a single connection is a canard.
However, there is a cause/effect relationship with the latter. They can't multi-home to multiple providers because they aren't big enough (can't justify the cost). Which is precisely part of the problem that I am presenting here.
> > What I am bringing up here is that new, information-age companies,
> > as predicted in MegaTrends over 10 years ago, are now starting to
> > appear. They are very diffused (sparse population, over very large
> > areas of the globe) and have connectivity needs which are
> both critical,
> > yet very different from click-n-morter customers that the Big8 was
> > built up to handle (either classful or classless). The
> current architecture
> > is not handeling them very well.
> > The problem is currently in it's infancy, it will get much worse.
> I'm not disagreeing with any of this. Actually, I see
> reliability and
> availability feeding into all these other issues as well.
The reason this is an issue is exactly because they want reliability and availability, HA requirements.
> It just that some of the folks advocating portability and
> deaggregation are
> using "route table size doesn't matter anymore" as an
> argument, when it
> absolutely does matter, especially if we plan to make the
> Internet more
> reliable, and less vulnerable.
I actually agree with you here as well. relying on infinite router table growth is not a scalable strategy. We need something else.