Re: Final word re: final word on pseudo-header

[Francis Dupont <[e-mail]>]

 Perhaps it is more useful to clarify some obscure points in the
draft-ietf-tuba-clnp-02.txt document (I suppose it is still the last one ?).

 There is a strange "Reserved" field after the source address in the
proposed Internet header using CLNP (Figure 4-1. CLNP for TUBA).
The text gives a false detail about the length of the addresses.

 In Table 5-1. Comparison of CLNP Error Reports to ICMP Error Messages:
----------------------------------|------------------------------------
Destination Unreachable     (128) | Network Unreachable         (3, 0)
Destination Unknown         (129) | Host Unreachable            (3, 1)

the difference between "Destination Unreachable" and "Destination Unknown" is
not so clear (:-) and "Host Unreachable" is safer than "Network Unreachable".

ICMP "Protocol Unreachable" (3, 2) is in fact a "Destination Unknown" case
(reported by the destination ES) because the (IP) protocol is the (CLNS) NSEL.

The ICMP "Port Unreachable" (3, 3) is very useful for UDP then I propose
to map it into a "Destination Unknown" with a second octet of the discard
reason option (which localizes the first octet of the erroneous field)
pointing to the destination port field in the UDP header (i.e. equal to
CLNP header length + 3).

We need an up-to-date description of pseudo-headers with examples showing
where are the paddings. Perhaps the best solution is to recommend the
Keith Sklower's hack (precompute the checksum on the length+address
structures and use the standard IP pseudo-header with the checksums at the
place of the IP addresses. There is only one difficulty: the source checksum
must be byte-swapped if the length of the CLNS destination address is even.
Of course checksums are cached...).

[e-mail]