Re: [I18ndir] I18ndir early review of draft-ietf-dispatch-javascript-mjs-07

[Asmus Freytag <[e-mail]>]

On 5/9/2020 11:36 AM, Barry Leiba wrote:
>> It occurs to me that http clients can send an Accept header, so how about
>> saying that modules SHOULD be text/jsmodule, but MAY be text/javascript
>> for backward compatibiltiy with clients that don't handle text/jsmodule?
> I'm very much not fond of the SHOULD... but MAY construction, because
> I think the MAY contradicts the SHOULD.
>
> I think the right way to say this is more along the line of "SHOULD
> use text/jsmodule.  If that is not practical because backward
> compatibility with clients that do not support text/jsmodule is
> necessary, then the alternative MUST be text/javascript."

Barry's formulation possibly implies that there's some third alternative 
that SHOULD / MAY allows without mentioning it. However, if that is the 
case, the new formulation allows this third alternative in every case 
where the implementer deviates from SHOULD for any reason other than the 
one given in the MUST phrase.

MUST be one of...

SHOULD not use text/javascript unless ...

This way the limitation to the two possible forms is explicit and the 
grounds for the exception are attached to a SHOULD phrase, which 
conceptually does imply the need for a justification.

You could also quibble that a MUST phrase that has an implementer 
defined "out" is against the spirit of the thing.

A./