W3C home > Mailing lists > Public > ietf-http-wg@w3.org > October to December 2021

Re: Media type parameters optional between semicolons

From: Carsten Bormann <cabo@tzi.org>
Date: Wed, 13 Oct 2021 15:05:55 +0200
Cc: ietf-http-wg@w3.org
Message-Id: <8DAF5EB8-C071-44E8-91CD-1F55473C5253@tzi.org>
To: Julian Reschke <julian.reschke@gmx.de>


> On 2021-10-13, at 14:52, Julian Reschke <julian.reschke@gmx.de> wrote:
> 
> Am 13.10.2021 um 08:56 schrieb Carsten Bormann:
>> On 2021-10-13, at 08:48, Julian Reschke <julian.reschke@gmx.de> wrote:
>>> 
>>> Am 12.10.2021 um 23:48 schrieb Carsten Bormann:
>>>> https://datatracker.ietf.org/doc/html/draft-ietf-httpbis-semantics-11
>>>> says:
>>>> 
>>>>    media-type = type "/" subtype parameters
>>>>    parameters = *( OWS ";" OWS [ parameter ] )
>>>> 
>>>> https://datatracker.ietf.org/doc/html/draft-ietf-httpbis-semantics-10
>>>> says:
>>>> 
>>>>    media-type = type "/" subtype *( OWS ";" OWS parameter )
>>>> 
>>>> Is the rationale for making the parameter optional documented?
>>>> 
>>>> Where -10 would have allowed:
>>>> text/plain;format=flowed
>>>> … 11 now allows:
>>>> text/plain;;;;format=flowed;;
>>>> 
>>>> Very curious.
>>> 
>>> See <https://github.com/httpwg/http-core/issues/33>.
>> 
>> Thank you!
>> So my conclusion is that this is on the same level as OWS or obs-text and should not be promoted in new specifications using this syntax.
> 
> I would characterize it as something that many implementations decided
> to accept despite it being invalid, and thus it got used in the wild.

Sure, trying to outlaw these implementations doesn’t work for HTTP.

But we are creating a different wild:
Using media-types (content-types and content-codings, actually) in SenML.

https://www.ietf.org/archive/id/draft-ietf-core-senml-data-ct-05.html#name-abnf (*)

> Not sure about what new specifications you have in mind. If it really
> replicates Content-Type syntax, I would just align with the current
> spec. 

Which is exactly what we don’t want to do — carrying that 25-year old baggage (HTAB, obs-text…) into new software that doesn’t have to (directly) interoperate.
(The use case really isn’t about copy-paste from an HTTP header to a SenML record; in that case carrying the baggage might make at least some sense.)

[CDDL has a way to identify syntax that is deprecated; maybe that is needed for ABNF.  But that might be for another list…]

Grüße, Carsten

(*) Yes, there is one bug…  Fixing that in today’s interim.
Received on Wednesday, 13 October 2021 13:06:11 UTC

This archive was generated by hypermail 2.4.0 : Wednesday, 13 October 2021 13:06:13 UTC