Re: [Idr] Question about draft-haas-flowspec-capability-bits-02

[Jeffrey Haas <[e-mail]>]

Linda,

On Mon, Jun 21, 2021 at 08:51:41PM +0000, Linda Dunbar wrote:
> Jeff,
> 
> One more question:
> Page 3 first paragraph says that "this is due to the lack of a mandatory length element for the components in the NLRI".
> 
> But Section 4.2.1.2 of the RFC8555 Figure 3 shows that the Len can be encoded in the bits. Are you talking about a different "length"?

If you observe the various forms of "length" in RFC 8955 section 4, you'll
see that in some cases it's an explicit one octet length (prefixes) mixed
with numeric op or bitmask op.  The latter two forms provide for a variable
length component type with variable width values.

If I introduce a component 14, how do I know what scheme it uses for length
calculations?  

The implementation must KNOW by the spec.  It is not carried in the
protocol consistently.  That is the problem.

-- Jeff