Re: [Cbor] John Scudder's Discuss on draft-ietf-cbor-network-addresses-10: (with DISCUSS)

[John Scudder <[e-mail]>]

Hi Michael,

Some comments in line below.

> On Oct 7, 2021, at 1:52 PM, Michael Richardson <[e-mail]> wrote:
[…snip…]
> I have changed it to say:
> 
> +would be parsed in the exact same way, they MUST be considered invalid by decoders.

OK. Maybe even a little better would be the same but append “(see below)” or “(see Section x.y)”? It just makes me twitch that you’re putting a MUST for *decoders* into a section titled “*Encoder* Considerations for Prefixes”. From that point of view the MUST is misplaced in this section and the text could be “… would be parsed in the exact same way, Section x.y. requires they be considered invalid.” 

> and also:
> 
> -A particularly paranoid decoder could examine the lower non-relevant bits to determine if they are non-zero, and reject the prefix.
> +Decoders MUST examine the lower non-relevant bits to determine if they are non-zero, and reject the prefix if they are not.

That definitely makes it more consistent. There remains the contradiction between that requirement, and the requirement right above, to coerce the low bits to zero. There’s no point in coercing them to zero if you MUST reject the prefix, is there?

So, ISTM that you need to either

a) Make the text quoted above a SHOULD instead of a MUST, move the “Finally, looking at the last three bits” text and pseudo-code to follow, and make it say something like “if the decoder chooses not to reject a prefix whose lower non-relevant bits are nonzero, then it MUST force them to zero…” 

b) Or if you keep the “reject the prefix” text as a MUST, then remove the coerce-to-zero text and pseudocode, because it’s the equivalent of a region of code that can’t be reached.

Thanks,

—John