>as i have been saying for a year or two. this bit of stupidity makes
>anycast useless for the majority of uses to which it is put today [0].
>i presume it will be ignored, and hence die when this stuff tries to
>go for draft. if not, a buch of v6 use will die [1].
regarding to anycast topic...
currently-practiced IPv4 "anycast" is done via injecting the same
address prefix by BGP from multiple locations - like the following
documents.
draft-ohta-root-servers-01.txt
draft-ietf-dnsop-shared-root-server-01.txt
the above approach assumes the following:
- session lifetime is relatively short - like DNS and http
- routing change is not frequent - at least, it won't happen
during session lifetime
- content provided by anycast server is identical
in the context of RFC2460, we cannot assume the above two - for
example, people may try to establish long-lived TCP session to anycast
address, and get RST suddenly due to routing changes.
so i think it reasonable for RFC2460 to have limitation in source
address selection, for general use anycast - it is a safety measure.
if RFC2460 recommends something like "anycast address can be used
only for short-lived sessions", it looks weird.
itojun
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
