[CFRG] Re: Where should test vectors live?

[Frank Denis <[e-mail]>]

Hi,

For implementers, JSON files are indeed be more convenient than the human-readable ones.

In the draft-irtf-cfrg-aegis-aead draft, the test vectors are included in the document as in RFC9180, but also available as JSON files at https://github.com/cfrg/draft-irtf-cfrg-aegis-aead/tree/main/test-vectors

The JSON files are automatically extracted from the document, so we can ensure that they always match.

That being said, the link to these files has been tagged as having to be removed by the RFC editor before publication, as this repository may not be suitable as permanent location.

Or could it? Would it be possible for GitHub repos within the CFRG organization to be considered stable URIs (as in: can be included in RFCs) to store test vectors and reference code?
That would greatly simplify things, both for document authors and for implementers.

-Frank.

> On Sep 20, 2024, at 02:17, Christopher Patton <cpatton=[e-mail]> wrote:
> 
> Hi CFRG,
> 
> It occurred to me today that our drafts often have human-friendly test vectors that look something like this:
> https://datatracker.ietf.org/doc/html/rfc9180#appendix-A.1.1
> 
> These aren't super convenient for a machine. In theory you could write a script that downloads https://www.rfc-editor.org/rfc/rfc9180.txt and write a parser to pull out the test vectors, but does anyone really do this? Luckily for RFC 9180 we have a JSON version to work with instead:
> https://raw.githubusercontent.com/cfrg/draft-irtf-cfrg-hpke/refs/heads/master/test-vectors.json
> 
> How do folks feel about pointing to machine readable test vectors from an RFC in lieu of producing human-friendly, but machine-unfriendly in the appendix? Suppose for example an RFC had a pointer to a JSON blob somewhere on datatracker. Is this feasible/desirable?