Re: Proposal: (node ...) construction rule type

Subject: Re: Proposal: (node ...) construction rule type
From: David Megginson <ak117@xxxxxxxxxxxxxxxxxxx>
Date: Thu, 14 Aug 1997 13:30:13 -0400
Paul Prescod writes:

 > I'll leave that as an excercise for the reader. You are right that it is
 > certainly in the "Scheme" (and DSSSL) tradition to describe language
 > constructs in terms of more primitive constructs.

It's more than just a tradition -- it's an important (and required)
sanity check that accomplishes at least two goals:

1. Make certain that the language is logically and completely
   described and implemented.
2. Avoid creeping featurism.

Your proposal is useful because it provides a fundamental mechanism
for describing the higher-level construction, rather than simply
adding another higher-level construction to the list.

 > > The problem is that there is no mechanism in DSSSL for adding
 > > construction rules, the way that there is for adding flow-object
 > > classes and external procedures.  If James modifies Jade to include
 > > the 'node' construction rule, I suspect that his program will no
 > > longer be DSSSL-compliant (at least, until the standard changes).
 > > Then again, he could provide a command-line option to disable the
 > > extension for full DSSSL compliance.
 > James has other DSSSL extensions. As long as he can accept a standard
 > stylesheet Jade would be DSSSL compliant.

Yes, but all of Jame's other extensions can be declared in
DSSSL-compliant ways, using 'external-procedure',
'declare-characteristic', or 'declare-flow-object'.  There is no
corresponding 'declare-construction-rule' procedure in DSSSL.  I don't
know how much of a problem that might be for James.

 > Anyhow, I think that all of the likely DSSSL implementors are on this
 > list. If there is a big flaw in the proposal then they will probably say
 > so. If not, they can implement it. As you point out, it doesn't seem
 > difficult.

With Jade, at least, it's almost trivial -- SP is already providing
all of the required information (although optimising the character
handling would take a bit of work).  From what I've seen, anything
based on NXP should have no problem either.

All the best,


David Megginson                 ak117@xxxxxxxxxxxxxxxxxxx
Microstar Software Ltd.         dmeggins@xxxxxxxxxxxxx

 DSSSList info and archive:

Current Thread