Re: [media-types] Thoughts on suffixes, single and multiple

[Mark Nottingham <[e-mail]>]

> On 6 Apr 2024, at 23:24, Manu Sporny <[e-mail]> wrote:
> 
> Is it worth, as a next step, documenting the current design patterns
> without determining whether they are "good" or "bad" yet? We could
> start with:
> 
> * Syntax suffix pattern (+json, +xml)
> * Container suffix pattern (+zip, +jwt)
> * Multiple suffix pattern (vc+ld+json, did+ld+json)
> 
> I can take an action to do that if folks are supportive.

I think that's workable, and will try to help. I suspect we're going to end up with text something like:

- '+gzip' MUST be the last suffix in a media type that uses it, and the results of unzipping it MUST be a valid instance of the format identified by removing the '+gzip' suffix. That media type MUST also be registered.
- When '+gzip' is present, fragment identifiers are resolved against the decompressed payload, using the resolution scheme of that media type.
- Any suffix used to the left of '+json' MUST identify how to serialise its specific format into JSON.

And so on.

What we need to sort out now, I think, is what implications this has for the multiple suffixes draft. I tend to think that it supports the 'bag of tags' approach that I outlined earlier; that puts the fewest constraints on how suffixes are defined (so that they can define their own constraints, as per above).

We also probably want to explicitly list what things a suffix registration needs to document.

Cheers,

--
Mark Nottingham   https://www.mnot.net/