Re: VCC cost models .... (was Re: Limits on SVCCs)

["Eric W. Gray" <[e-mail]>]

Grenville,

You wrote (in part):
> 
...
> Let's not get too carried away with the detail :-)
> 
> The 'service' is not TCP or UDP from this users's perspective. It's
> a web page, a RealAudio connection, an mpeg movie snippet. For
> these things I have a personal and somtimes dynamic sense of their worth.
> If you can collapse trade-offs like TCP vs UDP, or hop-by-hop vs cut-through,
> into some sort of comparative dollar figure, I can tell you which option
> I think is 'worth it'. Right at the moment, the axis along which most
> people will find cost sensitivity will be the long-vs-short distance
> ATM call. Which NHRP impacts on.
>

Do you consider yourself a typical user?

What about those cases where ATM attachment point is an edge device?  Do you
think it's a good idea to introduce user-feedback loop complexity between the
edge device and the non-ATM host(s) it represents?  Would you propose that a
pop-up window should appear on a network manager's work-station in such cases?

Do you imagine that any user wants to be bothered with a pop-up window (or
other "confirmation" query) for every potential shortcut connection?

My point was that the right place to make the decision to allow for a short-
cut is most often in the network (not the host).  A very likely approach to
doing this is through policy decisions made by network nodes (whether these
are switches, routers or NHSs) and - for those instances where dialogue with
a user is desirable - the policy may be to seek confirmation from the user.

> 
> Interesting problem, no?
>

Only as an implementation detail...

> 
> cheers,
> gja
> 
> (for the last year or so the related question of cut-through MARS
> has popped up a number of times. needless to say this, umm, makes
> for interesting back-of-the-envelope calculations as you figure
> out who pays when _you_ request to join a Class D group and suddenly
> force a number of senders to add a long-distance leaf node to their
> pt-mpt VCs.)

It may be due to my somewhat faulty intuition, but I suspect that _you_
(whoever _you_ is) pays...