[ticker-dev] Timeout attribute

[David Arnold]

-->"Matthew" == Phillips, Matthew <Matthew.Phillips at dsto.defence.gov.au> writes:

  Matthew> We should add a note about org.tickertape.message:3000
  Matthew> differences, especially the change in Timeout from minutes
  Matthew> to seconds.

I've added the following text to the section on Timeout:

    Note that this value was changed to use seconds, rather than
    minutes, quite late in the standardisation process.  To minimise
    the impact of this change on deployed applications, the following
    guidelines are provided:
    
    For any message advertising a protocol version of 3001 or greater,
    the value SHOULD be assumed to be in seconds.
    
    For messages with earlier versions, or that do not specify a
    version, Timeout values of less than 60 or equal to 1440 MAY be
    interpreted as a number of minutes.
    
    Similarly, implementations MAY generally refrain from sending
    messages with timeout values of less than a minute, and for a
    timeout of one minute, SHOULD use a value of 61.

Does that sound ok?


  Matthew> * We've gone back to "Attachment"? Not that I mind, but
  Matthew> didn't we have an exchange which resulted in going from
  Matthew> Attachment to MIME-Attachment a while ago? Also, using
  Matthew> string disallows some MIME encodings. I'd be inclined to
  Matthew> allow opaque or string for this field.

I have been in favour of using MIME-Attachment as a more specific name
to more precisely reflect the value's semantics.

It was my impression, however, that the consensus was that a better
solution was to use more general naming, and rely on collective
agreement on the semantics (via an ontology, data-dictionary, etc).

So some time ago I'd changed the spec back to Attachment, thinking
that it reflected the general mood of the community.

So speak up now if you favour MIME-Attachment!  Anyone?


Thanks Matthew,




d