[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: shim - transport/network communication



And I worry much about a solution which, for lack of thinking about such things, picks the worst path simply because it has less AS's in it or some such thing. We do far too much of that in the Internet.


On Mar 21, 2005, at 2:35 PM, Joe Abley wrote:

Hi Fred,

I am aware that I have much mail to catch up on in this folder, and I may well be missing your point, but this caught my eye:

On 21-Mar-05, at 10:08 AM, Baker Fred wrote:

I would like the multihoming facility we develop to have the characteristic of always picking a pair of addresses, and therefore a route, that gives among the best available RTTs (note I didn't say "the best"; I want to avoid the worst and accept a reasonable choice) to the current instantiation of the exchange.

I can think of lots of applications for which that might give the worst possible outcome.


What about an application that really wants the minimum packet loss, or the minimum jitter, or the maximum throughput? Or a path where 80/tcp isn't transparently redirected, or a path with no NAT, or a path where 161/udp is permitted? How about a path with a path MTU greater than 1500 bytes?

I would like any new multihoming facility to be deterministic (so that troubleshooting is possible), tunable (so that the ISP, or the user, or the application, or someone is capable of making reasonable adjustments) and stable (so that positive feedback loops are avoided). This does not necessarily spell "automatic" to me, and I don't think that's necessarily a problem.

I worry about talk of a multi-homing solution which automatically chooses "the" best path from a set of candidate paths for all traffic which flows through an AS.


Joe