Re: Resolving geo discussions

[Sean Doran <smd@ab.use.net>]

Noel -

On Monday, Apr 14, 2003, at 16:59 Europe/London, J. Noel Chiappa wrote:

> Also, this may lead ISP's to decide that they'd rather not support 
> exchanges,
Too late.  It has been many years since there has been any exchange in
the world with anything close to universal participation.   The 
paragraph
at the end of your message is thus not imagination but clairvoyance.

> since the customers are less locked-in
No, there are bigger problems ranging from whether the exchange operator
is vaguely competent now (let alone dealing with scale) to which of 
several
local exchange operators you want to pay to connect to (they aren't 
charities,
and they compete in places like London, Frankfurt and Paris), to which
of my customers is really likely to announce the exchange-related 
address
space in its entirety to me, given the trickiness of recovering the 
money they
pay me for the traffic this attracts from their competitors who also 
attach to
the exchange point.

If you believe in customer lock-in, then you are so out of touch with 
the market
in competitive places like Europe that I'd like to introduce you to 
several
people who would love to sell you their phenomenally successful due to 
customer
lock-in PA-address-using ISPs at low low extra value prices.  :-)

Site renumbering *is* a pain, but it's a pain a site might only have to 
do once
or maybe twice annually, given typical transit contract lengths, and 
there are
various ugly technologies that make the job easier.   (If I mention 
NAT, though,
I fear someone will seriously bruise his knees...)

Back to your neck of the woods, i.e., market-free theory :-)  -- what 
do you
think of three axes to describe the problem we're looking at:  the 
frequency
of the start renumbering, the  amount / clustering of renumbering 
things,
and the time to complete the renumbering process?    This could describe
a site changing providers annually, a laptop switching between different
wireless  networks,  or a "Schroedinger's network" that is 'stuck' in a
long renumbering process that ends only when a network connection 
fails, and
begins again when the connection recovers, among other things, and
has been an interesting thinking instrument for me.

	Sean.