On Wed 2017-05-31 08:45:34 -0300, David Bremner wrote:
> Not all keys will necessarily be generated with a given version of gmime.
> ---
>  devel/schemata | 33 ++++++++++++++++++++++++++-------
>  1 file changed, 26 insertions(+), 7 deletions(-)
>
> diff --git a/devel/schemata b/devel/schemata
> index 00ebb7a6..30f6ac53 100644
> --- a/devel/schemata
> +++ b/devel/schemata
> @@ -14,7 +14,7 @@ are interleaved. Keys are printed as keywords (symbols preceded by a
>  colon), e.g. (:id "123" :time 54321 :from "foobar"). Null is printed as
>  nil, true as t and false as nil.
>  
> -This is version 2 of the structured output format.
> +This is version 4 of the structured output format.
>  
>  Version history
>  ---------------
> @@ -30,6 +30,9 @@ v3
>  - Replaced message.filename string with a list of filenames.
>  - Added part.content-disposition field.
>  
> +v4
> +- replace signature error with a set of flags mirroring gmime 3.0 / gpgme signature status
> +
>  Common non-terminals
>  --------------------
>  
> @@ -120,17 +123,33 @@ encstatus = [{status: "good"|"bad"}]
>  sigstatus = [signature*]
>  
>  signature = {
> -    # (signature_status_to_string)
> -    status:         "none"|"good"|"bad"|"error"|"unknown",
> -    # if status is "good":
> +    flags:          sig_flags
> +    # if flags["good"]:
>      fingerprint?:   string,
>      created?:       unix_time,
>      expires?:       unix_time,
>      userid?:        string
> -    # if status is not "good":
> +    # if flags["bad"] || flags["error"]:
>      keyid?:         string
> -    # if the signature has errors:
> -    errors?:        int
> +
> +}
> +
> +sig_flags = {
> +    # exactly one of the next three will be set
> +    good?: bool,
> +    bad?: bool,
> +    error?: bool,
> +    valid?: bool,
> +    key-revoked?: bool,
> +    key-expired?: bool,
> +    sig-expired?: bool,
> +    key-missing?: bool,
> +    alg-unsupported?: bool,
> +    crl-missing?: bool,
> +    crl-too-old?: bool,
> +    bad-policy?: bool,
> +    sys-error?: bool,
> +    tofu-conflict?: bool
>  }

I understand that this is mapping to underlying gpgme/gmime status, but
the "exactly one of the next three will be set" business makes me
uncomfortable as an API.  what if it's not true?

If we've got a bitfield, we should expose it as a bitfield.  but if we
know that there is additionally a status component that has one of
exactly three values, we should expose it as its own element.

So i'd prefer:

  status: "good"|"bad"|"error",
  flags: sig_flags,
  # if status is "good":
   …

and then excise good? error? bad? from sig_flags.

what do y'all think?

     --dkg