Re: [core] Pub/Sub update to use the core multipart content format

Hi Jim,

I am proposing to use the multipart response to indicate that there aren't data to be returned on a subscribe operation.

We were proposing a new response code, but there was a lot of pushback on doing this, so we want to try the multipart format with an empty array (cbor x08).

It does require that a pub/sub client learn a little cbor and be able to unpack the intended payload from the embedded string.

Would the new response code be a better option after all? It it easier to teach clients about a new 2.xx response than a new format?

Best regards,

MIchael

> On Mar 2, 2019, at 7:41 PM, Jim Schaad <ietf@augustcellars.com> wrote:
> 
> Michael,
> 
> When I was suggesting using multipart I was thinking more in term of
> publishing rather than retrieval.  It would allow for multiple formats to be
> sent to the pub/sub server which the server would not need to think about
> doing format conversion on but could still return only a single format.
> 
> I don't think that one wants to do a multipart retrieval as a general rule
> unless the content really a multipart thing.  I guess that the server would
> need to be told which way to treat things, but that seems to be a reasonable
> query parameter for topic creation.
> 
> Jim
> 
> 
>> -----Original Message-----
>> From: core <core-bounces@ietf.org> On Behalf Of Michael Koster
>> Sent: Saturday, March 2, 2019 3:24 PM
>> To: Ari Keränen <ari.keranen@ericsson.com>; Jaime Jiménez
>> <jaimejim@gmail.com>
>> Cc: core@ietf.org WG <core@ietf.org>
>> Subject: Re: [core] Pub/Sub update to use the core multipart content
> format
>> 
>> OK, I read the multipart draft (expired?) For the no data case we will
> just
>> send an empty multipart payload (cbor x80).
>> 
>> It seems like every implementation of pub/sub will be required to support
>> only link-format and multipart. The subtype support will be on a topic
> basis
>> and maybe a broker should have some way to list a set of supported
>> subtopics that it can handle. To remind people, a coap pub/sub broker
>> doesn't have any notion of "resource state" other than a stored
>> representation.
>> 
>> Best regards,
>> 
>> Michael
>> 
>>> On Mar 2, 2019, at 5:27 AM, Michael Koster
>> <michaeljohnkoster@gmail.com> wrote:
>>> 
>>> Hi,
>>> 
>>> I am editing the pub/sub draft today and thinking about the multipart
>> content format as a way to provide a subscribe or retrieve response for
> the
>> no data case.
>>> 
>>> It seems to require that the pub/sub client always support the multipart
>> format, with no content being one of the format options signaled by
>> something like "maybe".
>>> 
>>> Does this seem right to you also? Do we need to invent a new empty
>> format code instead of a response code?
>>> 
>>> Best regards,
>>> 
>>> Michael
>>> 
>>> 
>>> 
>>> 
>> 
>> _______________________________________________
>> core mailing list
>> core@ietf.org
>> https://www.ietf.org/mailman/listinfo/core
> 