RE: COMPLEXITY BIG ISSUE

[David Burdett <david.burdett@commerceone.com>]

Christopher

I think we completely agree. We definitely don't want a "specification
language" the good old printed word (plus a few diagrams) is fine for what
we need to do right now.

I also agree that we specify it as a DTD first, but we should do it in
anticipation of the existence of a schema language.

What this means is that when we specify (in words) the structure of our
message headers, we should include Schema concepts such as datatypes and
inheritance (i.e. this element is like some other one, but with extra bits)
as this will make the transition to schemas much easier.

Right now in Commerce One, we have tools that allow you to translate from
SOX (our own XML Schema language) to a DTD, Microsoft's XDR or the current
draft of the W3C schema so going from one definition language to another one
is largely a mechanical exercise.

Remember though, that in the XML **instance** they should (or is it must)
all be the same.

David

-----Original Message-----
From: Christopher Ferris [mailto:chris.ferris@east.sun.com]
Sent: Monday, March 06, 2000 8:24 AM
To: David Burdett
Cc: David RR Webber; Dave Hollander; [unknown]; [unknown]
Subject: Re: COMPLEXITY BIG ISSUE


David,

Possibly I misunderstood your original point. I agree that there's a benefit
in keeping specific syntax out of the picture while we do our analysis and
design
as we are currently proceeding.

I was referring to whatever it is we choose to use for the formal
specification

as opposed to wasting our efforts on defining or selecting some syntax
neutral
specification language.

The idea of a style-guide is a good one, assuming we can find one. I don't
think
we should invest any effort on development of one. Possibly the Requirements
WG

or one of the other groups might consider spending cycles on this.

Cheers,

Chris

David Burdett wrote:

> >>>I think that we can skip the "syntax neutral" <<<
>
> Not so fast.
>
> There are another two reasons to go the syntax neutral appraoach first and
> leave the development of the DTD until a little bit later:
> 1. We will postpone (we can't avoid) the religous wars over where we use
> elements and where we use attributes, and how we name things, for example
do
> we have "<MessageHeader>, <MsgHdr>, <message.header>, <messageHeader> or
> some other permutation you can think of
> 2. It will force us to focus on analysis and design before we get to
coding
> ... how often do you see a DTD that has no documentation and you are left
to
> **infer** the meaning of an element or attribute from its name only.
>
> I think actually that the first point on use of names in an XML style
thing
> and wouldn't it be nice if there was an ebXML "style guide" for writing
XML,
> or does someone know a good one we could consider adopting
>
> David
> PS perhaps I'm starting the "religous war" by raising this issue now ...
;-)
>
> -----Original Message-----
> From: Christopher Ferris [mailto:chris.ferris@east.sun.com]
> Sent: Monday, March 06, 2000 6:34 AM
> To: David Burdett
> Cc: David RR Webber; Dave Hollander; [unknown]; [unknown]
> Subject: Re: COMPLEXITY BIG ISSUE
>
> David/All,
>
> I think that we can skip the "syntax neutral" aspect and use DTD for now.
> There
> exist tools to assist in the mapping of DTD's to Schema (albeit other
schema
> proposals
> such as SOX and XDR) and I'm confident that there will be tools to do the
> same
> with
> the formal W3C Schema recommendation once it is approved.
>
> I think that it is key that we go with what we've got and not be dependent
> upon
> that which isn't ready for prime time.
>
> One of the key drivers for ebXML is lowering the barriers to entry for the
> "little guy"
> which means that (IMHO) we need to provide standards which can be
> successfully
>
> implemented using readily (and inexpensively!) available tools and
> technologies.
>
> Everyone has access to validating XML parsers (SAX and/or DOM) today. Many
> are
>
> free. One cannot say the same for W3C Schema processors.
>
> Any cycles spent on selecting (or worse yet, inventing) some "syntax
> neutral"
> specification
> language will be lost towards the actual work required of this group, the
> actual definition of (at least)
> the 3 key deliverables you yourself suggested in a previous email.
>
> Cheers,
>
> Chris
> David Burdett wrote:
>
> > David says ...
> > >>>I'm very happy with 4 to 6 months [for W3C schema], seeing this
meshes
> > well with the ebXML timetable<<<
> >
> > I think that we can completely separate any dependency between the
schema
> > recommendation and our work if:
> > 1. We specify the data requirements and structure of any message
headers,
> > envelopes etc, in a way that is "syntax neutral", ie. we define a
> > hierachical structure of field names and descriptions where all the
field
> > names are expressed in natural english or short phrases.
> > 2. We map the hierachical structure to relevent XML definition languages
> > specifically a DTD and a W3C Schema if it's available and judged
> > sufficiently stable.
> >
> > Thoughts?
> >
> > David