Phelan, Tom wrote:
Hi Gorry,
See inline for comments on comments :-). Sorry it took so long :-(.
Tom P.
[snipped]
I added a ref in 3.2:
NEW:
Network address and port translators, known collectively as NATs
[RFC2663], may interpret DCCP ports [RFC2993] [ID.Behave-DCCP].
I also updated bullet 2 to explicitly refer to this:
NEW:
o A middlebox that does not modify the intended application (e.g.
NATs
[ID.Behave-DCCP] and Firewalls), MUST NOT change the Service Code.
[Tom P] Is this a hypothetical situation -- are there any examples of
middle boxes that *change* one application into another, different
application? Or am I misunderstanding what you're saying?
We spoke in one meeting about RTP-relays and such like, but as I recall
we decided that it would not be a good idea to go down this rat-hole of
defining specific actions for such things - just to say NATs and
firewalls MUST NOT change the SC.
I have currently made this a normative reference - since this is a PS
that speaks of a specific mechanism that is important for SC use and
dedicated to DCCP, but I could make it informative if you think this
better reflects the citation.
[Tom P] Normative reference sounds right to me.
Done.
Section 3.3.1 (mostly editorial, but maybe technical):
"If the receiving host is listening on a server port and the
DCCP-Request uses a Service Code previously associated with the
port" --
Would "uses a Service Code *currently* associated with the port" be
clearer? The current wording seems to suggest that any SC that was
ever
associated with a port is OK.
Aha, I now suggest:
OLD:
uses a Service Code currently associated
NEW:
uses a Service Code that has been associated
[Tom P] That wording seems to me to say the same thing -- that any SC
ever associated with the port should be accepted. How about "uses a
Service Code that *is* associated"?
OK.
Changed to /is/ (and one similar clause also set to present tense).
Section 3.3.3 (technical):
So what should be the identifier in dccp-inetd? Just a port or just
an
SC or a port and an SC (the last, for my thinking)? I think that
should
be specified here. I think you try to specify this in the last
paragraph, but it seems unclear to me.
OK, well spotted - this is vague, but intentional. I hedged away from
detail on the inetd discussion (after various iterations with Eddie
and
Mark). I do not have much to say from the OS point of view - I guess I
was thinking of driving this from some of file/database with SC and
port.
[Tom P] So is there an action here?
I do not have anything to add.
Gorry
