Date: Wed, 8 Oct 1997 19:14:21 -0700
From: "David L. Wasley" <David.Wasley@ucop.edu>
To: Elise Marie Meyer <emm@physics.ucsb.edu>
Cc: cenic-tpg@ucdavis.edu
Subject: Re: CalREN-2 Query

Elise, I don't operate a campus network but I'd like to add some thoughts
for your consideration.

QoS is not a solved problem - not even close.  Not only are there serious
issues about what it means but there is a need for admission control for
which there is not any serious model yet (that I know of).  Therefore, I
would suggest waiting for clear dierction before investing in equipment.

One aspect to consider is "scope": over what segment of the transport do
you want the QoS to apply?  For example, if the problem is not on campus
but between MAE-West and Denver, then you really want to pay only for
enhanced service on that segment.  In practice, this might devolve to
enhanced QoS only within a given administrative domain - i.e. an ISP
backbone or a campus or ...

Put another way, "end-to-end" is a misleading notion.  What you want is a
minimum quality level of Internet transport service between cooperating end
nodes.  This can be achieved in a number of ways and need not require
enhanced service along the entire path.  In fact, it is likely that there
will be proxies at each border crossing that will accept and validate
requests so that the application can request (and pay for) enhanced service
only where needed.

	David