Re: [mpls] Alexey Melnikov's Discuss on draft-ietf-mpls-rfc3107bis-02: (with DISCUSS and COMMENT)

[Alexey Melnikov <[e-mail]>]

Hi Eric,

On Thu, Aug 17, 2017, at 02:51 PM, Eric C Rosen wrote:
> H Alexey,
> 
> In BGP, anything that is announced by a Capability is always optional.  
> Per the draft, if one doesn't send the Multiple Labels Capability, one 
> sends exactly one label in the NLRI.   So in that sense, the mandatory 
> minimum value is 1.  I thought this part was clear from the first bullet 
> point in Section 2.

Yes, this part was clear.

> When the Multiple Labels Capability is sent at session establishment 
> time, I thought the new text:
> 
> "A triple of the form <AFI=x,SAFI=y,Count=0> or <AFI=x,SAFI=y,Count=1>
>   MUST NOT be sent.  If such a triple is received, it MUST be ignored."
> 
> implied that the mandatory minimum value is 2, as that text precludes 
> the use of the only smaller values.

I got that. I thought 2 was a rather low limit, but you made a good
explanation in this email, so I am Ok with that. So I will clear my
DISCUSS.

> If you think this isn't clear enough, I can some text like "Any 
> implementation that sends a Multiple Labels Capability MUST be able to 
> support at least two labels in the NLRI.  However, there may be 
> deployment scenarios in which a larger number of labels is needed."
> 
> Would that be okay?

Probably not needed, but I am Ok if the above is added.

> I wouldn't want to make the mandatory minimum any larger, because (a) 
> some deployment scenarios might only require two labels, and (b) we 
> don't want to have a larger mandatory minimum because the number of 
> labels is not merely a control plane issue, but also affects the data 
> plane.  E.g., if some platform can only push two labels at a time, 
> there's no value to requiring it to support three labels in the control 
> plane.
> 
> Eric
> 
> 
> 
> 
> 
> 
>