Hello Bill,
There's a big difference between allowing refusal (which I agree should be possible) and requiring support in the implementation. I can make an analogy between your example and the IPsec mandate. Your example "works", because there aren't any communications that last for very many packets. What if I build a product that ONLY ever supports insecure applications? Should I be allowed to claim IPv6 conformance if my product platform does not have any IPsec implementation? Currently, the answer for IPv6 is "no". But, I reckon the number of nodes that only ever support insecure applications would be a lot more than the number of nodes that only ever support instantaneous, disconnected application streams. Plus, one could even devise semi-rational scenarios where the DNS server needs to support mobility, of course depending on the applications. For remote diagnosis, maybe some (secure!) streaming applications for processed or graphic data would be highly advantageous, and the diagnostician might be mobile at the time... In fact, this doesn't sound at all far fetched to me. Regards, Charlie P. Bill Sommerfeld wrote: > > Here's another reason why RO needs to be optional: > > Consider the traffic patterns surrounding a root name server. > > Properly functioning clients talk to it at most once every two days > (the current TTL of root server NS records) per TLD used, and do a > single packet exchange per TLD. > > There would be no point in it ever maintaining a binding cache. > > On the other hand, I don't see much harm in mandating that hosts be > able to politely refuse a binding update request. > > - Bill -------------------------------------------------------------------- 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] --------------------------------------------------------------------
