I will not respond to your other message as you seem to have a fundamental misunderstanding of the context of the word "send" in the rate-limiter specification, so any discussion about that would be useless.
See the clarification below: On Tue, 17 Aug 2004, Alex Conta wrote: > Please read the specs before making these statements. I would ask you to do the same, and taking the context into account. > The text is clear in the ICMP specs: "...limit bandwidth and forwarding > costs..." > > Please read paragraph (f) page 7 of the ICMP specs, which says: > > (f) Finally, in order to limit the bandwidth and forwarding costs > incurred by sending ICMPv6 error messages, an IPv6 node MUST > limit the rate of ICMPv6 error messages it sends. > > This points me to say that you must be confused, misinformed, or grossly > misinterpreting. If you look at the whole section 2.4, it is blindingly obvious that the word "send" is used to refer to the node *generating* (i.e., originating) an ICMP error message (in response to some other packet doing something illegal), *NOT* in the more generic meaning which could include forwarding an ICMP error packet it has received from some other interface. Do you agree with this? If you don't, ... I wouldn't me amazed as this seems to be your fundamental misunderstanding. If you do, and re-read the whole section again keeping that in mind, and (f) in particular, you should be able to tease it apart: - "in order to limit the bandwidth and forwarding costs *incurred by* sending ICMPv6 error messages" Note the word "incurred". At least in my book, this does not require that these forwarding/bandwidth costs happen on the node that is generating the ICMP packet. In other words, the algorithm is specified to reduce the number of ICMPv6 messages which backbone routers etc. will have to forward over the Internet, not so such the bandwidth of the generating node (note that there isn't even a "forwarding cost" of generating an ICMPv6 message, it's generation and sending cost). [the byproduct, of course, is that the node originating ICMPv6 errors doesn't waste all of its resources sending huge amounts of ICMPv6 errors] - "an IPv6 node MUST limit the rate of ICMPv6 error messages it sends." Note the word *sends* and its context. This specifies that ICMPv6 error messages must not be *generated* without limiting. This specifies nothing of routers *forwarding* received ICMPv6 messages. ... Do you agree? Whether you agree with this or not, I guess we'll need to reword some of the text to avoid confusing ICMPv6 error *generation* rate-limiters to *generic* (ICMPv6) rate-limiters (which you keep referring to) which are completely out of scope of this specification. Removing pretty much everything relating to bandwidth could probably be a good start, and using the word "originate" instead of "send". -- Pekka Savola "You each name yourselves king, yet the Netcore Oy kingdom bleeds." Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
