In the past we have
considered the initiator of IP transactions to be irrelevant and had
no-charge peering for networks that basically send a similar number of
bytes to what they receive.
So what do we do when that is no longer the case?
Pray someone doesn't write a call-back protocol, so traffic flows in
the opposite direction from the original initiator of the IP 'transaction'.
That becomes popular in a webbrowser. Or a popular game.
Then you'd have to monitor more information than just total traffic flow
in and out of an interface.
For video and voice over IP type services, the traffic flow should (largely)
be symetrical. But what else would be that makes this issue complicated?
Right now the name of the game is content. Email, web, news, realaudio,
realvideo. Later on when interactive (ie symetric traffic) applications
like voice/video over IP come into play, wouldn't it make more sense to
handle 'interconnecting' between two different voice/video services by
some other way?
(Please, if I've missed something here, I'd be glad to know...)
Puzzled,
Adrian
FTP already does this. Nevertheless, the largest number of bytes still
flows in the same direction as it would with HTTP. I think you are
agreeing with me that the initiator of the transaction is irrelevant. I
would go further and say that the customers of the peers and their actions
are also irrelevant to the peering relationship.
> Pray someone doesn't write a call-back protocol, so traffic flows in
> the opposite direction from the original initiator of the IP 'transaction'.
FTP already does this. Nevertheless, the largest number of bytes still
flows in the same direction as it would with HTTP. I think you are
agreeing with me that the initiator of the transaction is irrelevant.
Michael, you've always struck me as one of the saner inhabitants of
this list -- which I guess really translates as "you and I almost
always have the same outlook on things" 
-- but this must be where
we part company.
In the current context, which I would translate as "who is responsible
for the bytes moving over a link -- and therefore ought to pay for it",
it's pretty obvious to _me_ that if Exodus' customers are sending data
to GTEI's customers _because the latter requested it_, then Exodus
ought not, in equity, to be considered "responsible" for that data;
they were just doing as asked.
I would go further and say that the customers of the peers and their actions
are also irrelevant to the peering relationship.
On this, hoewver, I agree. The real breakage here is GETI attempting
to redefine "peering". The net got where it is today as a
"non-settlement" network. Any plan to change that would have to be
documented in about 50 pages for me to buy it.
Cheers,
-- jra