Fixed in git. We parse 'type' as a valid field, but still generate 'object-type'. This is for backwards compatibility reasons.

My concern for that issue is that URLs generated by gnutls will not be readable by other applications and with the resolution at hand this is still the case. Wouldn't it be more natural to read 'object-type' in addition to 'type' and generate 'type'? Compatibility with previous versions is not a major issue as URLs are not transmitted over systems, and the downgrade of existing packages is not something that is expected to work seamlessly (e.g., you cannot downgrade shared libraries after you have compiled applications with the new one).

Great idea. I would merge a patch for that right away.

Created attachment 111209 [details] [review]
Generate the type=value instead of object-type by default

Thanks. To summarize:

    We continue to accept both the older style 'object-type' field
    in addition to the new 'type' field. However we start generating
    URIs in the new form.
    
    In other words we have backwards compatibility, but not forwards
    compatibility. Given the fact that PKCS#11 URIs are now standardizing
    this is an acceptable compromise.

Pulled into git master.

(In reply to Stef Walter from comment #6)
> Pulled into git master.

Thank you. The PKCS #11 draft is already accepted (it is in the RFC editor queue): https://datatracker.ietf.org/doc/draft-pechanec-pkcs11uri/