[media-types] "Unwrapping" structured suffixes

[Jeffrey Yasskin <[e-mail]>]

Hi media-type folks,

When looking through the existing structured suffixes in
https://www.iana.org/assignments/media-type-structured-suffix/media-type-structured-suffix.xhtml,
there seem to be 2 kinds: ones like +xml, +cbor, +json, etc., which
describe the meta-format of the media type they appear in, and ones like +gzip
and +zstd, which appear to say that you can uncompress a foo/bar+gzip media
type to get some bytes that you can then process as foo/bar.

In my reading,
https://www.ietf.org/archive/id/draft-ietf-mediaman-suffixes-07.html#name-media-types-with-multiple-s
reinforces this possibility by referring to the left side of the first + as
the "base" subtype. I wouldn't expect to need a "base" media type if
something weren't unwrapped into that base format.

However, based on a discussion with some of the proponents of the
mediaman-suffixes draft in https://github.com/w3c/vc-data-model/issues/1364,
I think this interpretation is not what the authors were hoping to get
across. It sounds like +gzip and +zstd are also expected to be just
descriptions of the overall media type, and a tool isn't supposed to unwrap
a layer of suffix and try again with the "base" media type.
https://www.ietf.org/archive/id/draft-ietf-mediaman-suffixes-07.html#name-partial-processing-of-suffi
mentions a security problem with this possibility but doesn't directly say
not to do it.

I don't have a strong opinion about what do with this, and I don't want to
unnecessarily delay the RFC, but Harald thought it would be useful if I
posted to describe this confusion.

Jeffrey