GMPLS TLV Format

["Adrian Farrel" <adrian@olddog.co.uk>]

Hi,

In today's meeting, we discussed again the issue of TLV formats in:

- RFC 3209 and RFC 3471
- RFC 4420
- draft-ietf-ccamp-ethernet-traffic-parameters-03.txt

The current position is:
- Base RSVP-TE and GMPLS carry the full length of the
 TLV in the Length field
- OSPF carries the length of the Value field only in the
 TLV
- RFC 4420 follows the OSPF form. This was an error
  that was not intended.
- draft-ietf-ccamp-ethernet-traffic-parameters followed
  RFC 4420 (assuming that the WG had made a deliberate
  change)

The bottom line is that it is not helpful to implementers that one protocol 
has two different ways to encode TLVs.

We must choose between three options.
1. All TLVs are encoded as per RFC 3209.
   RFC 4420 would need to be fixed.
  draft-ietf-ccamp-ethernet-traffic-parameters would
   need to be changed.
2. All TLVs *except* those in RFC 4420 use the
   RFC 3209 format. RFC 4420 remains an anomaly.
3. All old TLVs remain as per RFC 3209. All new
   TLVs (starting from RFC 4420) use the RFC 4420
   format.

The meeting seemed to prefer option 1, but this is contingent on existing 
implementations. If there are too many existing and deployed implementations 
(too many == 1 ?) we may have to pick to option 2.

So...

How would you feel if we did an update to RFC 4420 that fixed the TLV 
encoding to be conformant with RFC 3209? Would this cause any of you a 
problem?

Reply on-list or to me in private if there is a confidentiality issue.

It would be helpful to have opinions on both sides. Please don't leave 
silence to mean anything specific.

Thanks,
Adrian