I agree. jak
----- Original Message ----- From: "Mark Andrews" <[EMAIL PROTECTED]> To: "Fred Baker" <[EMAIL PROTECTED]> Cc: "Brian Haberman" <[EMAIL PROTECTED]>; "James Kempf" <[EMAIL PROTECTED]>; "Margaret Wasserman" <[EMAIL PROTECTED]>; "Bob Hinden" <[EMAIL PROTECTED]>; "IPv6 WG" <ipv6@ietf.org>; "Joe Abley" <[EMAIL PROTECTED]> Sent: Wednesday, April 06, 2005 9:50 PM Subject: Re: Anycast support in draft-ietf-ipv6-addr-arch-v4-02.txt > > > OK, so then lets go back to the question posed in the thread. The > > current spec says that one should never use an anycast address as a > > source address under any circumstances. That clearly flies in the face > > of present practice, isn't responsiv to the set of concerns you raised > > about anycast in general, and can be mitigated if anycast is use for > > rendesvous. The suggestion was made that under a defined set of > > circumstances (single message each way exchange, rendezvous, perhaps > > some others) and with a defined set of procedures it would be OK to use > > it as a source address. Those procedures need to spell out the whys and > > wherefores. > > > > Would you be willing to see the 100% ban removed from the draft > > standard and substitute text to the effect of that above included, with > > follow-up work in grow-or-wherever to spell out those procedures? > > Definitely yes. > > > On Apr 6, 2005, at 8:19 PM, Mark Andrews wrote: > > > > > > > >> > > >> On Apr 6, 2005, at 6:36 PM, Mark Andrews wrote: > > >> > > >>> Getting back to unicast initiated sessions I would still > > >>> like to see some mechanism (as low in the stack as possible) > > >>> which would allow long running session to survive routing > > >>> changes. > > >> > > >> You're speaking in this thread. Did you take a look at the proposal > > >> that Eric Nordmark, I, and the grow folks have discussed about a > > >> care-of-address that would give a long term fixed address to the > > >> server > > >> in question? Answering that question is where we started out. > > > > > > care-of-address would be overkill for somethings and > > > quite a reasonable solution for others. If it could be > > > made selectable on a per/socket basis (I havn't looked > > > at how implemetation do this at present) I suspect this > > > will meet most of what would be required. > > > > > > In other words we would not want to do this for DNS/UDP but > > > for DNS/TCP it would be acceptable even though it would only > > > be really required for long running AXFR's (multi-megabyte). > > > > > >> -------------------------------------------------------------------- > > >> IETF IPv6 working group mailing list > > >> ipv6@ietf.org > > >> Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 > > >> -------------------------------------------------------------------- > > > -- > > > Mark Andrews, ISC > > > 1 Seymour St., Dundas Valley, NSW 2117, Australia > > > PHONE: +61 2 9871 4742 INTERNET: [EMAIL PROTECTED] > > > > -- > Mark Andrews, ISC > 1 Seymour St., Dundas Valley, NSW 2117, Australia > PHONE: +61 2 9871 4742 INTERNET: [EMAIL PROTECTED] > -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------