Re: [mpls] [bess] draft-rosen-mpls-rfc3107bis

[Eric C Rosen <[e-mail]>]

On 4/1/2016 4:23 PM, Robert Raszuk wrote:
> Hi Eric,
>
> I have read your proposed draft as well as watched this thread with a 
> bit of an interest.
>
> To me the best compromise - which is to agree with Bruno's points as 
> well as address your intentions is simply to request new SAFI for 
> 3107bis.

I don't think that makes any sense at all.  The whole point is to ensure 
that 3107bis interoperates with 3107 for those features that are already 
deployed and are already multi-vendor interoperable.

>
> From the draft you are really not updating 3107 base spec but 
> obsoleting it which to me looks like a bad idea.

That is the nature of a "bis" draft.

>
> You are even requesting to remove IANA reference to original spec.

That is the nature of a "bis" draft.

> How would IANA know when is it safe to do that .. meaning when all 
> implementations will not suddenly support and all deployments will 
> enable 3107bis ?

I don't understand the issue you are raising.  I don't see any issue of 
"safety".

>
> New SAFI requires a new capability which you are asking for anyway.

I don't understand the point you are making here.

>
> As far as implementations please keep in mind very important point 
> that some implementations treat SAFI 1 & 4 in single table and some in 
> separate tables.

Yes, and these implementation differences have consequences that are 
discussed in the draft.

> That when mixed with 3107bis may just explode if not in new set of 
> bugs then with operational nightmare.

I don't understand the issue you are raising here.

> While we are at this it would be much cleaner to mandate in the new 
> spec to have 3107bis always to use separate tables as compared with 
> from SAFI 1.

Two goals of this draft are (a) to document the consequences of the 
implementation differences, and (b) to avoid invalidating any particular 
implementation.  Obviously these goals would not be met if the spec 
mandated a particular implementation method.

> As we all know 3107(bis) tries to add NNI to MPLS. However it must be 
> very well stated that this is only one deployment option for 
> interdomain encapsulation. I would very much like to see a section 
> indicating that IPv6 or/and IPv4 be used as an alternative encap for 
> those applications which require it and when needed provide local 
> bindings between intradomain MPLS and interdomain IP.

This is entirely out of scope.