[e2e] traffic engineering considered harmful

Panos GEVROS P.Gevros at cs.ucl.ac.uk
Wed Jun 13 13:40:47 PDT 2001

"David P. Reed" typed :
 |At 05:01 PM 6/13/01 +0100, Panos GEVROS wrote:
 |>"David P. Reed" typed :
 |>    Economic history is
 |>  |littered with ventures that lost dominance because they loved control too
 |>  |much.  Look at GM, Xerox, Kodak, France.
 |>of course this applies to endpoints and "intermediaries" alike.
 |>(especially when the first dogmatically refuse to surrender -some- the 
 |>(freedom of action) the posses in the original Internet architecture)
 |If there were (or ever was) one or more "United Endpoints" that exerted 
 |control, I'd be more sympathetic to this point.  But it seems like a clever 
 |fallacy.  A category is quite different from an organized economic actor.

the "United-Endpoints" would be the isp for all its attached endpoints
(if i understood correctly the comment)
that is the scope in which "some" control will be surrendered

 |Endpoints seem to be being asked, as a class, to give up freedom of action 
 |for no benefits whatsoever.  Did they do anything to harm (economically) 
 |the ISPs that grew rich by making them happy?

it would not be just for giving ISPs peace of mind, not for restoring trust 
for its own sake, certainly not. and although this email is not a cost benefit 
analysis ;
the benefit would be the maintainance of end-to-end design (since it is not 
intelligence that is pushed inside, but responsibility) by removing many of 
the considerations which have put this design under threat.

by removing the considerations of the selfish-endpoint imho would lead in a 
completely different way of thinking about many problems in areas from 
resource sharing, congestion control, security etc.
today there may be far too many things happening in a network for which 
responsibility belongs to an endpoint (the broken protocol implementation it 
has and other nasty things) , because the endpoint is the master of its own 


More information about the end2end-interest mailing list