RE: intertwined specs

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

ASP is ugly when one first looks at it, not because it 
doesn't work, but because the combinations of all of the 
syntaxes plus quote counting and the awkwardness of 
syntax-sensitive editors *fixing things* are maddening. 
I consider it a bridge technology useful as other 
technologies coalesce. XQuery is a result of that 
coalescence.

The formal model is good for those who have to ensure that all of 
the specs in total remain coherent.  Perhaps the DEPH is being 
left behind and that is a shift of target audience for XML. Being 
a tool user, not a tool builder, I am not concerned but I can 
understand that individuals who thrive on tool building might 
be until they master the formal model.  Then I suspect, like 
some loathe BNF until they master it, it becomes the description 
of choice.

SGML works without LL(1) because the implementors hide that 
from the end user.   SGML is nicely flexible past that task. 
As long as the specs keep the pain of the implementor from 
the user, I'm not too concerned.  I concur with Marcus that 
we should not force them to master our solutions.  One 
odd and not really good side effect of HTML was to convince 
a large group of people that they were programming.  There 
really are levels to this that most here, I'm sure, understand.

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: Jonathan Robie [mailto:Jonathan.Robie@SoftwareAG-USA.com]

At 03:52 PM 2/16/2001 -0600, Bullard, Claude L (Len) wrote:
>As I sort my way through the XQuery draft, it has a
>way of looking like an ASP page with all of the
><% %> delimiters removed.

Yes, I think all of this is reasonable. Of course, many, perhaps most 
XQuery queries will not look like an ASP page. Many will just use the 
expression language without constructing literal XML.

>MSL:  well, ok as long as these are formal
>proofs among interested parties as to the
>correctness of parts, no problem.  If this
>enters the production systems, it is not
>something most will welcome outside the
>deepgeeks with the faster propellors. So,
>it will be an academic exercise similar to
>the early DSSSL with Scheme/LISP.  Good
>to have for formal work; not something
>the production folks will want to work with
>unless it is made *very* friendly by tools.
>Even then...

You don't show your formalisms to your users, no more than you ask them to 
read the assembly code.

>Keep in mind what someone made very clear
>very passionately in Vancouver when this
>SGML On The Web thing was a'bornin' and
>people had to agree to walk away from the
>advanced state of extant solid work:
>
>"It has to be simple."  Jean Paoli- Microsoft

Well, yes, it has to be simple. But formal approaches, used well, make 
things simpler. I wish SGML were LL(1), for instance, it would have made 
the language much easier to implement. However, I wouldn't bother my users 
with an explanation of what LL(1) means.