On 11/07/2016 07:13, George Michaelson wrote:
> Things like client capability signalling, I suspect are in a harder > bucket. I won't say intractably hard, but last time I floated > capability flagging, I got pretty strong pushback. I think capability flagging is difficult in a stateless protocol like UDP - there's no guarantee that the next endpoint that you send a packet to is the same one that just told you what capabilities it has. It could be done over TCP, and perhaps in other protocols that add a pseudo-session on top of UDP (e.g. DNS-o-DTLS?) but even then the signaling should only really apply to the current connection / session. As it happens, the Session Signaling draft I submitted last week originally had an option for exchanging the end-points' EDNS0 option support, but some of my co-authors didn't like it so I removed it... Ray _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
