Re: Favorites (Re: UUNET peering policy)

[Danny McPherson <danny () ambernetworks com>]

> Honoring MEDs is not Black and White. I dont know one large network that
> honors all MEDs from every peer. For the most part, I would say that most
> providers do 'warm' routing. it just depends on what degree of warm you
> can live with.

Of course, the amount of brokenness involved with honoring 
MEDs is questionable anyways.  For example, you can't compare
MEDs from different ASs and attain any reasonable result.  

For that matter, the result may very well be persistent route 
oscillation.  

And often enough, I've seen MEDs result in worst-exit routing. 
That is, routing that was intended to be better than closest-exit
(aka. hot-potato), but turns out to be far worse than best-exit
(I believe some refer to this as cold-potato *8^/) because an 
adjacent AS in unable to correctly convey the optimal entry point 
into their AS via MEDs, be it because of GOOD aggregation, or 
uncluefully derived MED values.

Then again, I've seen MEDs work just as intended, allowing a peer 
network to easily manipulate route/forwarding decsions in a 
neighboring network in order to alleviate congestion or other
issues within their own network.

Of course, there are clever ways to make MEDs of benefit, though
IMO, they do more harm than good overall.

-danny