Closed gorryfair closed 9 months ago
This is a standards document. Middleboxes that compute UDP checksums based on IP lengths are objectively misbehaving and incorrect in their computation. This isn't a research assessment. If it belongs anywhere, it belongs in a spec affected by that decision, IMO. We previously called them misbehaving and errant. They're wrong in what they do. Is there a preferred term for such that DOES NOT imply that they're not flat-out wrong?
I.e., could we NOT avoid being judgmental about protocol standards violations?
I'd persoanlly prefer to just say middleboxes that do not comply with RFCxxxx. That's more polite and helpful than saying "errant" or "misbehaving".
Polite to whom? Middleboxes don't have feelings or practice etiquette. ;-) Although I see your point, we need to use stronger language. Failure to comply is one thing, but these middleboxes create non-interoperability. It's not like they're setting UDP==0 in situations they shouldn't - which is failure to comply, but might not have an impact.
I thought the word "errant" (which was used before) was more elegant than "misbehaving" 😃
Errant is fine to me; I can unify around that term if useful.
I agree: Errant with respect to the spec seems a statement of fact!
Changed all occurrences in -24 to "errant"
This was agreed at the Interim.
I personally do dislike the words chosen here : “The design enables traversal of misbehaving middleboxes that incorrectly compute the UDP checksum over the entire IP payload”. To me, it is great for a research paper, but awkward for a standard because it sounds like a judgment on the way others do things, well maybe it is. Could we we can find more positive text and say what it does?
Such as “This design enables traversal of a middlebox that has been designed to compute the UDP checksum over the entire IP payload”. This also applies to later use of "misbehaving middlebox"