Re: KISS (was: Parameter entity references in WF docs)

Paul Prescod wrote:
> 
> len bullard wrote:
> > Unless the XML requirements are clear, adding more and more SGML
> > features
> > makes XML into SGML and only by sleight of hand gets rid of a perception
> > some here on this list and in other places have worked hard to foster:
> > SGML is Too Hard and XML is Simpler.
> 
> XML is still VERY far from the implementation complexity of SGML. Do you
> feel you have are very familiar with the SGML standard? 

I had to digest that standard when it was still a draft and 
you were playing with power ranger toys.  Luckily I learned 
SGML from implementations, not the standard.  That's clue 1.

> Would you be
> willing to undergo a little "quiz" administered by Erik Naggum? 

:-)  Sorry, Paul, but I've given up hitting my head with a 
wooden board everyday for pennance.  Suggest you do the same.

> Would
> you be willing to write a parser that implements all of the features?

Nope.  Nor will anyone else on this list.

> Think about LINK, minimization, subdoc, asyncronous entities, etc. 

I used to think of little else.  I got better.  However, I never 
found those hard to use except for LINK which I've never seen 
used.  The lore about it is deep enough to keep most people 
away.

> It is good to argue that feature X makes XML too hard, 

For whom?  In what context?

> but I think that
> arguing that it makes it "almost as hard as SGML" is wishful thinking
> about the complexity of the SGML standard. Remember: SP doesn't
> implement all of SGML yet.

It isn't the complexity of the SGML standard I'm referring to.
You'd have to implement HyTime as well to get that.  So far, I 
know almost everyone who has done that.  No thanks.  I'm 
really not worried about XML implementors.  Never have been.
They are paid to write code.  Hope they sell a lot of it.
The mythical CSG is a myth.  May he remain a hit or myth.

I'm worried about XML users.  They are real and if they 
aren't we are wasting time here.  XML is throwing a lot of stuff 
at them SGML never did because SGML didn't have it unless 
you count HyTime.  XML will take considerable chops to use.
 
> So now we are arguing about how far along the spectrum of complexity
> (implementation time) XML should go: 5% or 10%. 

Implementors?  PEs reduce the workload of the DTD designer and 
that is about it.  When someone implements all of XML-LINK, 
first I want a copy, then I'd like for them to quiz Erik Naggum.
I'd pay to watch that debate.  But a lot of good consulting 
dollars will be spent on XML experts PDQ because it isn't 
a simple markup language or even a simple hypertext application.
This is a document database language, not SGML or HTML.

> Some thought that XML
> was supposed to be just SGML without the hardly-used crap. Others
> thought that it was going to be a minimal extensible markup language. I
> think that we have been successfully doing both: applications that are
> not interested in validation have a very restricted set of features to
> deal with. Applications that want to do validations must know a lot more
> about content models, PEs, etc.

Yep.  I watched the last two phases of VRML development.  I watched 
some of the best and brightest computer scientists in the world 
design that language.  Keeping it simple wherever they could has 
paid handsomely.  Even then, it has been hard to get the applications 
to interoperate, and a lot of work to maintain even the simple 
content. Every good designer on that list 
knew the range of 3D design and rendering tricks.  They 
pled for features they loved.  Wiser heads prevailed.  It 
was a good thing they did.

Every feature was one more incompatibility until fully 
implemented in every browser.  To repeat, framework + plugin = browser.
The dimensions of complexity are complex numbers, not simple
percentages.
  
The problem of HyTime wasn't that it didn't work; the problem 
was no one could find out until it was done and it was never done.
I have the utmost respect and affection for it's designers and 
poured not an inconsequential amount of my own effort into the cause, 
but in the end, if it takes that long and it is that hard, a 
simpler scope of effort is demanded.  We live, we learn... we hope.
 
> If we restrict paramater entities to the *non-required* subset then only
> validating parsers will ever have to deal with them. Perl Hackers and
> Browser Vendors will not.

I don't disagree.  But when you start presenting this to users, 
not SGML designers, not DTD designers, not me or you or Martin or 
Deb or Eve, or the other 1/100000 of the community that has to do 
this first time, the fewer features the better.

So my rant is and remains, if you can live without it for a season or
two, 
dump it.  Get your software into the marketplace.  Find out what is 
needed, by the user (not us) then come back for XML 2.0.

len

Received on Monday, 2 June 1997 21:51:00 UTC