Re: New Version Notification for draft-troan-6man-universal-ra-option-03.txt

[Ted Lemon <[e-mail]>]

On Oct 9, 2020, at 8:28 AM, [e-mail] wrote:
> The motivation for this work was to "make 6man out of a job", since I thought we spent way too much time discussing configuration information options. :)
> 
> Do you have other concerns around this too Ted?
> It's hard to predict how this could be used of course. I suppose you could define an DHCP route option with this. But you'd still need an agent in the OS that would act upon that configuration. And if you had that, you could obviously also do that with whatever other protocol you'd like to invent.

I think your motivation is interesting, but you’re still going to need a WG to write documents describing new options in this new format—we still need interop.

The DHC working group actually wrote a document on how to encode TLVs because we saw the same problem; this predated CBOR, and I don’t know if we would have been tempted by CBOR had it been available. My parser in ISC DHCP had a little option definition language that was used to parse options. Not exactly CDDL, but same idea.

I think formal descriptions of options as you have done with CDDL is definitely a good idea. I’m less enamored of CBOR simply because it’s the wrong tool for the job—we have to do extra work because it does extra stuff we don’t need. I get that because there are CBOR parsers readily available, it’s the most attractive nuisance, and I am not necessarily saying “definitely not,” but as I’ve already expressed, I definitely have reservations about it. 

It would be interesting to think through what we get extra with CBOR that actually matters for this use case—I’m obviously highlighting the negatives, but perhaps there are positives I’m missing other than “there are already high quality parsers available.”  E.g., one advantage I can see for CBOR with a validator is that if the two ends don’t actually agree on the data format, you’ll notice the problem quickly, whereas if you were just parsing a binary blob it might be less obvious that you, e.g., got the fields out of order or something.

I think if you aren’t actually proposing to redo existing options in CBOR, it would be good to make that really clear. I will admit that I skimmed, so I might have missed your dire warnings to that effect.