BBN Peering issues (fwd), like all content hosting sites, attracts
users to one location, when it could instead attract users to
the same content in multiple locations.

This is inefficient.

Many archive sites direct people to mirrors topologically closer
to them, and as more actual content owners purchase duplicate sites
in places like Europe (,{se,, .ca}) both for the technical gains
and also as a means of strengthening one's brand globally.

This is much more efficient.

This should be carried on.

It is in carriers' interests to encourage the trend too,
and it is probably a good idea to work out means of hosting
content more locally than data centres in California,
as a means of increasing network efficiency.

It is in content-owners' interests to encourage the trend
also, as ultimately their brand is the one that is hurt
by unreliability. Yes, it might hurt carrier X's business
when carrier X cannot get to Popular Content Site, but some
fraction of carrier X's customers will go away thinking,
"performance to Popular Content Site sucks! Popular Company sucks!"

Moreover, as the capacity market distortions are sorted
out in many places, it almost certainly will be cheaper
for content owners to spread the work of distribution around,
so that the traffic stays as local as possible.


Many archive sites direct people to mirrors topologically closer
   to them, ...

Actually, most of the ones that do this with text the user is supposed
to understand, direct5 people to sites taht are GEOGRAPHICALLY
closer. Sometimes this is the right thing, and sometimes it isn't.
Schemes such as Exodus and others use, actually do produce the
topologically closer server, and that is, in fact, better...and
doesn't require the user to think (in this stage of growth, that's a
good thing :-).