Thu, 26 Jan 1995 01:52:26 +0100
|
On Tue, 24 Jan 1995 21:31:26 CST Bill Harvey <[log in to unmask]> said:
>Could the migration to TCP/IP take care of it? I guess I haven't given
>it much thought, but it seemed to me the biggest hurdle was that
>topological information wasn't (readily) available in the IP world. How
>will the DIST protocol fit into this "brave new world" of ours? (I'm not
>looking for commitments... Just a hint about where things might be
>headed...)
Oh, this is already implemented in 1.8a. It's not fully functional yet
because we don't have 200 LTCP servers yet, but the structure and code
are there. SERVICE NAMES maps users/service areas to servers and there is
some topological information in INTPEERS NAMES. Nowadays it is no longer
necessary to achieve 100% optimal routing. What matters is not to send
the same data 100 times across the same bottleneck. So to take a European
example, your campus gets one copy that it got from your national server
that got it from the US and you can, nowadays, claim that you have solved
the problem.
Eric
|
|
|