Re: What is a regional exchange point?

On Wed, 5 Mar 2003, Bill Woodcock wrote:

> 1) You're creating a category ("regional" exchange
> point) which presumably differs from an exchange point
> by its property of being "regional," yet you don't
> clearly delineate what distinguishes it as different
> from an exchange point in the usual sense.

I think the category existed long before I created the Utah
REP in 1995. There are quite a few other (North American)
exchanges that call themselves REPs or consider themselves
REPs. My intention isn't to create a new category, simply to
define one that apparently exists, at least in the minds of
some regional exchange operators.

I thought the delineation was implicit, in that a "regional"  
exchange focuses on interconnection within some particular
region, in the same way that "national" exchanges focus on
the ... nation.

The regional context also implies to whom it's important:
companies, municipalities, ISPs and users within Utah
consider CommIX to be very important, where a national
exchange is just a national exchange (do they really care if
it's PAIX or Equinix or a MAE, and if it's Chicago or
Dallas?) and a regional exchange in another state is
probably not important at all. 

Conversely, national exchanges are incredibly important (and
may be very distinctly different) to national-level networks
and users, who may consider regional exchanges
less-critical. Somewhere between there, lots of gray area
(national providers may peer at regional exchanges to
improve service into that region, and regional networks peer
nationally all the time).

Peering policies may be constructed differently for a
regional exchange in the US just like the are for a regional
exchange in Asia or Europe (eg. a national backbone may be
interested in peering Utah routes at CommIX, vs US routes at
EQCHI vs European routes a LINX).

I think there are some (more) obvious differences between
the focus, purpose, sponsorship and participants at most
REPs vs other exchanges. Maybe those need to be more clearly
spelled out in the definition (and I was trying to keep it
short).

> 2) Use of the word "regional" in this sense (or at least
> the sense I believe you're using it in) was common in
> the `80s, when there were "regional networks" but it's
> now evolved to mean something completely different, as
> in "regional internet registries."  This is unfortunate,
> as it confuses a lot of people who go back and read
> historical documents from before that shift of meaning,
> but I don't know that it can be helped, and not
> recognizing that change just confuses things further, in
> my opinion.

I think "regional" still means the same thing: some kind of
logical geographical grouping. Each region may have
different needs, justifications and motivations for an
exchange, and I don't profess to even try to understand the
complexities of regional peering outside North America (and
maybe not even outside the US).

The exchange exists because there are networks and users
within that region, whether it be a city or metro area or
state or multiple states, who want better connectivity
between themselves. Justifications may include reducing
local traffic latency (very popular in areas > 1000 miles
from a national exchange), lowering transit costs,
supporting economic development, keeping local traffic local
(Texas pride syndrome), pooling network costs, supporting
delay- or jitter-sensitive applications.

Pete.