There is an annoying edge case where a certificate with no DNS- or
CN-IDs can have a Redacted Labels extension with zero elements and not
be invalid according to 6962-bis.

This means you can't immediately reject a certificate if the Redacted
Labels extension has fewer than one element.  Combined with the
requirement that the last integer implicitly repeats, this made my
TBSCertificate reconstruction implementation more complicated than it
would be otherwise.  Considering that this code involves indexing an
array, which carries a risk of invalid memory access in memory unsafe
languages, I think it's important to make it possible to detect and
reject bad input as easily and as early as possible.

If redaction is not removed, could we specify that the Redacted
Labels extension MUST NOT be present if no labels are redacted?

Regards,
Andrew

_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

Reply via email to