Date: Mon, 3 Oct 94 09:03:50 JST
From: francis@cactus.slab.ntt.jp (Paul Francis)
Message-Id: <9410030003.AA00314@cactus.slab.ntt.jp>
To: J.P.Knight@lut.ac.uk, liberte@ncsa.uiuc.edu
Subject: Re: Multicasting vs Anycasting (was: No "TOP" of the docuverse)
>
> Has there actually been an anycasting service deployed on the Internet?
>
To my knowledge there has not. The network layer folks seem busy with
some other trivial matters (can't remember just what off hand).
There is also some debate among network layer folks about the utility
of anycast. Steve Deering, for instance, believes that multicast can
be made sufficiently efficient for use as a discovery tool (for instance,
through expanding-ring searches) that unicast is unnecessary.
I think we should experiment with anycast. I believe it may be possible
to implement anycast without any modification to the current IP
infrastructure. If you advertise the same unicast address from multiple
machines, the resulting behaviour should be that a packet sent to that
multicast address goes to the "closest" advertiser. I'm not sure what
effect route filtering would have on this, so it may not work. The
point is, it should be possible for someone to start experimenting with
anycast today....
PF