Re: [dtn-interest] UDP/DCCP draft

[Joerg Ott <[e-mail]>]

Hi,

please find below my review for the UDP/DCCP convergence layer
draft.

Overall, the document is in good shape and almost ready to go.

Comments:

Section 3.2: somewhere here or in section 3.1, we may want to
point out the overhead of using bundle layer fragmentation since
the per-datagram overhead may be quite substantial.  No normative
language, just as a hint.

Section 3.4: We probably want to specify guidelines to look at
for the lower/upper bound for keep-alive messages.  The IETF
recommends 600s for UDP traffic, but experiments have shown that
NAT boxes may have as short timeouts as 30s.

Moreover, if both ends support keep-alives, is there a procedure
to follow if they receive keep-alives (in certain intervals)?  Or
is this function implemented for each direction individually?
Either is probably fine, but it would be useful to state this
explicitly.

Section 3.6: "or it MUST": this probably is meant to refer to
the bundle protocol using the convergence layer, not the dgram cl?
Should just be made explicit.

Nits:
If section 4 remains empty, we can remove it.
Sect. 2, 3rd para, 1st line: it's -> its


That's all I got.

Jörg