[CFRG] Re: Where should test vectors live? // Suggestion with example draft

[Björn Haase <[e-mail]>]

Hi to all,

as I did spend some time just yesterday with preparing test vectors in
JSON format for the upcoming next CPace draft version, I have today
patched my test vector generation scripts for including JSON in the
original text body of the ID in line with my today's suggestion for an
answer:
Suggested answer to the question "where should test vectors for CFRG
drafts live?":

"
1.) The test vectors SHALL live in the RFC and/or ID document
2.) Test vectors SHALL be made available in form of a textual "human-readable" format using base16 encoding.
Test vectors SHOULD also BE provided redundantly in a second format that is machine-readable. It is RECOMMENDED to use a base64 encoding that
decodes to a JSON file as machine-readable format.
"

At

https://cfrg.github.io/draft-irtf-cfrg-cpace/draft-irtf-cfrg-cpace.html#name-test-vector-for-cpace-using

you see how this possibly could look like.

What do you think about the approach taken there ? Maybe have a glance
at section B.1. to B.1.1.1 . For decoding the base64 sections you could
use your preferred online base64 decoder tool.

With the approach taken there we would avoid the difficulty of
maintaining a long-term link to some server.

I think for the end-user this might be even more convenient than having
a separate link. If test vectors are anyway machine-generated by scripts
(such as for CPace) integrating test vectors as part of the text body is
not really a lot of a burden. Also you maintain the textual context
specifying which exact test vectors are included.
The price that we would be paying is that we would have to accept some
redundancy (and have to check two outputs) and a bit longer RFC documents.

As comparison you could also look at the compiled total test vector file
for the CPace draft at

https://github.com/cfrg/draft-irtf-cfrg-cpace/blob/master/testvectors.json

I think that in this file it might be not really straight-forward to
locate the right section in the large file that applies to a given
subsection of the ID appendix.

I also would appreciate feedback regarding your opinion on old-fashioned
C-language constant tables as in B.1.8. What do you think? Are plain C
test vectors useful or should still we rather only use more modern JSON
as single format instead and forget about plain old C?

Yours,

Björn.


--
Diese E-Mail wurde von AVG-Antivirussoftware auf Viren geprüft.
www.avg.com