It makes sense, but there is one pitfall. More and more, firewalls come with restrictive rules about what ICMP packets will be allowed. These rules typically operate on ICMP type and ICMP code. Having a single type-code combination for all experimental packets will force an all or nothing decision in firewalls, allow all experiments or deny them all.
The Code is defined on a per-Type basis, and in our proposed format, on a per Sub-Type basis, so, of course, we are not proposing (see my previous msg) a single Type-Code combination.
As for grouping *all* experimental protocols together, well, yes, that's the goal in order to save on ICMP Types. But the info to distinguish amongst them is in the fixed part of the header (the Sub-Type field) so any decent firewall can differentiate.
This may be OK if experiments are limited in time and in scope, but it may become a liability if experiments linger on -- and we can use Mobile-IPv6 as an example of an experiment that can linger on for a very long time.
Not sure what MIPv6 has to do with this. MIPv6 is not an experimental protocol, right? It's proposed standard. Perhaps you meant 'experimental' in the sense of not highly deployed, etc. But that would apply to a host of other IETF protocols as well.
Or, perhaps you're referring to the fact that MIPv6 assigned 4 new ICMP Types? If so, yes, it's this kind of consideration that I believe leads the AD's to want to be more conservative in the future.
...
This format basically adds 1 bit to the type field. I wonder whether that is enough. Why limit the subtype to 8 bits? Why not use 32 bits, and some decentralized allocation scheme?
Our goal has not been to change ICMP Type field semantics, so perhaps we have not even added the 1 bit you mention (from 7 to 8 bits for informational messages). So the idea is not to grow the ICMP Type space to a whopping large number. We hear it is best to be conservative in doling them out, but I don't believe the problem is serious enough to warrant growing the Type space significantly.
Data point: from a previous thread, the IPv6 chairs noted that the total number of assignments was:
ICMPv4: 27 ICMPv6: 23
So the situation is not desperate.
-gabriel
-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
