Fwd: Blocks versus Header and Body

        Here are a few thoughts after last telcon discussion
about whether we should distinguish header blocks and body blocks or
not.

        An issue was raised concerning this distinction and usage
scenario S21. If an XMLP sender is generating a lengthy message and
transmitting it incrementally, it will not be able to add a header
containing a checksum for the message.
        Mark Jones offered a solution to this issue by proposing to
put the message content in a header, to add another header containing
the checksum and to link the body to the first header.

        IMHO, this is a workaround, rather than a proper solution.
I am somewhat reluctant to accept it, as we have in our requirements a
whole section about simplicity and stability. In particular, I think
that both R307 and R308 call for simplicity, which to me translates
into removing the artificial distinction (I believe) between header
and body blocks.

        Moreover, if a resource constrained device receives a lengthy
message where the content is in a header, and parses it incrementally,
another workaround is needed to indicate at the beginning of the
message that the content is in a header. Otherwise, the device will
learn this only while parsing the body, when the header content has
already been discarded.

        As we have a requirement, R309, that says that we should care
about resource constrained devices, I think we should have a proper
solution for solving S21, rather than using workaround.

        So I think that we should keep this simplification of SOAP
in our minds and work on it in due time to enhance our specifications.

        Best regards,

        Hervé.

Received on Wednesday, 25 April 2001 03:16:02 UTC