In your previous mail you wrote:

here are two articles on dnsop by bill fumerola in response to this thread
there.

=> this shows the fall of T/TCP has left a hole...

since sctp is slightly older and is used in ip telephony and is
implemented on more systems and allows data transfer on the second round
trip (packets 3 and 4), i'm more hopeful for it than i am for DCCP.

=> what we want is data transfer on the first round (packets 1 and 2)
but it opens the door to fake source (in packet 1) attacks...
DCCP is younger but its "market" is far larger. My concern is no
new transport can really replace UDP and as a DNS fallback TCP.

i see five reasons why SCTP is worth looking at as a primary transport with
UDP as its fallback. first, data transfer can take place in the second
round trip (packets 3 and 4). second, it has 32-bit ID fields so we could
go back to fixed port numbers perhaps even fixed+increment on TXID's.
third, we could mandate EDNS, and with it, we could mandate PING. fourth,
SCTP has the idea of "passive" sockets which mean nonquadratic socket pool
sizes. fifth, association timeout is handled in the stack not the app.

=> IMHO this only shows SCTP has the pros and cons of a new transport.

Thanks

Francis.Dupont@fdupont.fr

--
to unsubscribe send a message to namedroppers-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: