> On Jul 23, 2019, at 11:19 AM, Theresa Enghardt <ther...@inet.tu-berlin.de> > wrote: > > Hi, > >> On 23/07/2019, 10:31, Tommy Pauly wrote: >>> […] if you don't require reliability, then a UDP protocol stack is >>> equivalent to a message protocol over TCP. If you do require >>> reliability, then those aren't equivalent. > > Another important difference between TCP and UDP are Message Boundaries. > So in some cases, TCP + Framer may be equivalent to UDP.
Exactly! That's what I meant by "a message protocol over TCP". > > > We probably don't want to put that in the Architecture draft, but a > Framer can essentially provide the "Message boundaries" property, thus, > making TCP equivalent to UDP if Message boundaries are required, right? Yes, I think that's already implicit in the text in Section 4.2.3, point 1. TCP doesn't preserve boundaries on its own/if it's the top (but does if there's another protocol that does). > > Maybe that's something for Implementation draft? But the application may > also want to know about the possibility to essentially use TCP or UDP > without having to change (much)? Thinking about DNS, even if there may > be more stuff to consider here, such as the EDNS0 features Ted Hardie > mentioned. I agree that the implementation document is a more appropriate place to discuss the particulars of which protocols can be determined to be equivalent. Thanks, Tommy > > Best, > Theresa > > _______________________________________________ > Taps mailing list > Taps@ietf.org > https://www.ietf.org/mailman/listinfo/taps _______________________________________________ Taps mailing list Taps@ietf.org https://www.ietf.org/mailman/listinfo/taps
