Re: Datatype section

[Joaquin Miller <miller@joaquin.net>]

Philip Goatly wrote:
> ...    I agree Betty has done
> an excellent job in surfacing all these data
> types.
>
>     BUT surely we must remember that any XML message only
> contains text
> data either say ASCII or UNICODE etc. depending on the XML 'ISO'
> settings. ...


Phil is right to commend Betty. Good work. And he goes on to make a very
important point.  

Perhaps the decision has already been made, either formally, or by
unspoken consensus, that ebXML needs to define these datatypes, even
knowing that they will be transmitted as text.  (I'm a recent
lurker; that's why i don't know.)

If not, then that needs to be discussed.

If so, then we clearly need to specify at least the text format for each
type.  And we need to discuss whether it is also necessary to
specify conversions to other representations. (Which other
representations?)   Probably everyone knows that already. 
Maybe you all agree on the answer.

Here are a few items that are a bit deeper than text format and
conversions.  Since the list comes from a draft standard submitted
to ISO, perhaps this is not the place to mention these items.  If it
is not, please delete this message now.

The list starts of innocently.  String and boolean {true, false},
are easy to let slip by.  Float and double are well defined by
IEEE.  Decimal can slip by, too (we imagine one through nine with
one period in there somewhere).  

But timeDuration opens up the dread box.  And there are some real
troublemakers in there.  Examples of minor ones are Integer, which
mentions 'scale' and a couple of xIntegers, which mention the even more
obscure 'maxInclusive.'  Those are probably defined somewhere.

timeInstant opens up another box found in the first.  I don't want
to go all philosophical on you, but what is an instant?  might we
not want to have a more realistic idea of time?  There is a whole
body of work on this topic, and it is vital to get this right for
distributed systems.  (I feel that the only realistic model is based
on intervals, not instants, but that is only my opinion.)

The concept, 'midnight,' is famously treacherous.  It's use here
seems ok, but is right at the edge of the cliff.  And it
is a cliff, not just a slippery slope.  There is a
rock solid reason lawyers never use the term, nor the time.  

Wait a minute!  I just read Date again.  Classic case of
midnight trouble.  Of course you all know what is meant, but how
about an arbitrarily selected computer program?

recurringDay and her sister, recurringDate, will be interesting.

And you will all join me in enjoying century.  Happy New
Century!  Or am i late on that?

Cordially,

Joaquin

................................................

Joaquin Miller

Chief Architect

Financial Systems Architects

mailto:joaquin@acm.org

San Francisco

phone: +1 (510) 336-2545

fax:   +1 (510) 336-2546

PGP Fingerprint:

CA23 6BCA ACAB 6006 E3C3 0E79 2122 94B4 E5FD 42C3