Pascal Thubert wrote: > Face it: Many of today's large web servers will not want to maintain binding > caches. Because: > - They are clustered (see my previous mail on this thread) > - Visits are short and unrelated (the caches are not reused) > - RR reduces the throughput and the response time
The current spec allows CNs to decline a RO/RR request. This is of course necessary, we couldn't mandate that they have sufficient memory etc. in all cases to handle these requests. > - They may even blindly accept the Home Address option since the AAA takes place > at session level anyway using cookies and such. Most likely we can't do this. Earlier analysis indicated that unverified home address option could lead to reflection attacks. > How much of the global IP traffic does this represent today? What will mobile > devices do at least in the short term? We must provide a way for servers to > simply and politely decline RO. We may want to permit triangular routing. The > minimum we can do to ease IPv6 transition is at least to allow the current v4 > applications to run in equivalent conditions. Yes, and they can decline RO already now. Triangular routing will only be allowed under a special condition (existing SA), which I think is probably not compatible with your scenario of web servers. > It's a SHOULD. I'm not really arguing against that... Jari -------------------------------------------------------------------- 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] --------------------------------------------------------------------