[e2e] It's all my fault

[Ken Calvert]

>> Allowing source routing at the level of transit providers
>> shifts the balance of power back toward the user. (See
>> Xiaowei Yang's thesis.)
>
> Are we living on the same planet?  Do you seriously think that any ISP
> would be interested in purchasing equipment or software which would let
> users to get the best of the "balance of power" (or, to put it more
> bluntly, ability to screw ISP's traffic engineering and business
> arrangements with peers)?

No, I don't expect this would be deployed by current 
providers.  It requires a fairly complete re-thinking of the 
architecture because other mechanisms are needed to make it 
work.  It will only be deployed if, as was suggested, it 
(i.e. a new architecture) grows in parallel and eventually 
succeeds because it is more attractive in some way.  It 
would take a long time, if it happens.

What's needed are mechanisms that separate concerns to allow 
the different parties in the "tussle" to implement their own 
policies. It is not obvious (to me -- but I'm neither an ISP 
nor an economist) that allowing competition among transit 
providers precludes traffic engineering a priori.

>> It's really about the interesting possibilities that cannot even be
>> contemplated because of the lack of such a mechanism (and others needed
>> to make it feasible).
>
> Oh, surely one does not need actual deployment to contemplate all the
> interesting possibilities.

You are right, one can contemplate.  But as soon as one 
starts talking about it, lots of folks with a firm grasp of 
the status quo start saying it'll never work, there's no 
market for it, etc.  The only way to overcome that is to 
build something and use it (what I think Reed was talking 
about).

> boring. There's a good enough technology which more-less works. Years ago
> I figured out how to build backbones with arbitrarily large capacity,
> there are no more technological challenges in that.  This market is all
> about price/performance now, a commodity market.

It's a commodity market because for the vast majority of 
customers, the denominator of price/performance is 
essentially fixed.

KC
-- 
Ken Calvert, Associate Professor      Lab for Advanced Networking
calvert at netlab.uky.edu                     University of Kentucky
Tel: +1.859.257.6745                 Hardymon Building, 2nd Floor
Fax: +1.859.323.1971                              301 Rose Street
http://www.cs.uky.edu/~calvert/          Lexington, KY 40506-0495