Re: [dhcwg] DHCPv6: Canonical way to handle an unknown option

[Ted Lemon <Ted.Lemon at nominum.com>]

This raises an interesting question - in several places RFC3115 talks  
about a "valid Reply message," but the term is never defined.   We  
aren't told what makes a Reply message "valid" or "invalid."   So I  
understand the writer's confusion, and this ought to be clarified.   I  
would assume that "valid" means "well-formed," which is to say "parses  
without encountering anything the RFC hasn't told you how to parse."    
You've been told how to parse options, so I would say that an unknown  
option is valid in this context, but the RFC provides no guidance.

Anyway, the bottom line is that logically it doesn't make sense to  
drop packets with unknown options - it's entirely possible that these  
options represent a valid protocol extension that was added after you  
implemented your client, and it would be a mistake for your client to  
fail because of this.   Rather, it should ignore what it doesn't  
understand, as long as it's well-formed.   Otherwise, some protocol  
enhancement down the road could suddenly result in your client failing  
dependably, and you really don't want those phone calls.