Re: [Rats] I-D: draft-rundgren-cote-00

[Michael Richardson <[e-mail]>]

Anders Rundgren <[e-mail]> wrote:
    > In this case my assumption is that the overarching goal is providing an
    > object typing mechanism permitting a common receiver dispatching the
    > handling of received objects to different processors depending on their
    > type.

I think that in 90% of the situations, the processor expects a very limited
number of types, often only one.

    > In RATS it appears that you are currently dealing with not less than
    > three different typing systems (media types, CBOR tags, profile URLs)
    > which (in my simple mind...) feels slightly over the top.

media-types, encoded by Content-Formats over CoAP apply to data in transit.
CBOR tags (whether via file-magic, or just alone) go with the data at rest.

I'm not sure where I'd situate profile URLs.
It just seems like lazy coding by people used to online software where you'd
better have the latest code, or you are screwed.

To me, they only make sense if there is active code at the URL that can make
sense of the content.
While I can imagine such a thing for something detailed, complex and never
standardized, it's hard for me to make up a reasonable example.
(Maybe there is a sufficiently rich declarative data description rather than
active code, but then that data description is itself a well known content)

If as you said in another thread, the software to intrepret the complex data
is already locally installed, then it must also be sufficiently well defined
that a trip through IANA isn't a problem.

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     [e-mail]  http://www.sandelman.ca/        |   ruby on rails    [