Hi Stewart, As for address embedding public key, it need not to carry any algorithm in the address. It would be much better to carry the public key by address, while indicate the algorithm by protocol. I think CGA is a good instance for involve address in cryptography. For forwarding efficiency, a public key can be only set as a suffix, thus forwarder could process the prefix only, and thus the cryptography related stuff may not hinder the looking up efficiency.
While for the prefix table lookup, method like Multi-Entrance-Trie (METrie) can be used to support fast table lookup (see https://dl.acm.org/doi/pdf/10.1145/3341558.3342204), the evaluation in this paper shows it is possible to achieve efficient packet forwarding for length variable IP address. Thanks, Yihao 发件人: Stewart Bryant [mailto:[email protected]] 发送时间: 2021年2月5日 22:59 收件人: Jiayihao <[email protected]> 抄送: Stewart Bryant <[email protected]>; int-area <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; Lin Han <[email protected]> 主题: Re: The small address use case in FlexIP On 5 Feb 2021, at 12:06, Jiayihao <[email protected]<mailto:[email protected]>> wrote: - Indeed, the network scale of limited domain is supposed to be less that IPv6, but it doesn't mean the address space should be strictly less than 128-bit. If the space of the address is abundant enough, the public key could be embedded without truncation (compare to CGA in IPv6) for certain security purpose. Interesting, what are the advantages in adding the signature of the address in the address as opposed to carrying it in a different field? The disadvantage is that you bind the address to the signature algorithm which you would not want to do since you would expect to change the signature algorithm during the lifetime of the protocol. Also would you really want to feed the signature into the longest match engine? Of course you could and there are some advantages in that you look up both the address and it signature, but I think you loose longest match capability and you significantly increase the size of the TCAM or other FIB design memory, and that memory is very expensive as it determines the line rate of the forwarder. So this points back to the need for a holistic discussion of what we are trying to achieve, the extent to which modifying existing protocols satisfies that need, and whether (given the presupposed need for a gateway) we should be looking for a single protocol, a family of protocols, or an adaptable protocol. I don’t think we can design the addressing system in the absence of a discussion on those points. Best regards Stewart
_______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
