[mpls] draft-bryant-mpls-sfl-control - Control Response Codes

[Stewart Bryant <[e-mail]>]

> On 3 Jan 2020, at 02:44, Zhenghaomian <[e-mail]> wrote:
> 
> 
>  
> 5. section 3.1 SFL Control Message
> Regarding the 'Control Code', do we allocate specific value for the query/response actions? For example, we can have request (control code = 0), refresh (control code = 1) and withdraw (control code =2) for a query. Currently there is no such allocations.

Thanks for picking this up. It is quite significant since without these values you cannot implement the protocol.

I brushed over this about five years ago when I quickly pulled the first draft together and so far no one (including the authors) noticed. The implications is that no one has implemented the protocol. Obviously I think it will work, but you cannot be sure with a signalling protocol, and as the WG knows I have been a bit nervous on that point. I wonder if we should downgrade it from PS to Experimental?

The codes need to go into a registry and we have two choices, we create a new registry for it or we recognise the string ties to RFC6374 and add the values to its registry (presumably updating the RFC and renaming the registry).

It is tempting to put the values into RFC6374 because of the strong heritage but I made a mistake like that before with the ACH code points and it caused all sorts of problems, so I will clone the RFC 6374 IANA text and use it to create a new registry. However it makes me wonder if we should be trying to get some commonality in the ACH

In doing that I note that we originally proposed 0x60. I think we meant the next one after 0x59 which is 0x5A, so unless anyone objects I suggest we request 0x5A rather than leave the gap in the registry.

I will compile and upload the latest version, but need to take a much loser look at the error codes and I think we can do a better job than I have done so far.

This will need at least one more revision, but so long as it is on track the chairs could adopt it and allow it to proceed under WG control.

Best regards

Stewart