Re: DSSSL Design Question

Subject: Re: DSSSL Design Question
From: James Clark <jjc@xxxxxxxxxx>
Date: Mon, 26 May 1997 12:34:03 +0700
At 23:55 25/05/97 -0400, Paul Prescod wrote:

>It would also be nice
>to be able to pass information "down" the flow objct creation tree more
>generally than characteristics allow.

Since you can declare your own characterstics, why aren't they sufficient?

I think it would desirable to allow #f instead of a public identifier in
declare-characteristic, to mean that this isn't a characteristic that has
semantics that should be passed to the formatter, but rather it's just being
used to pass information down.

Is it the fact that you can only get at characteristic values within the
specification of other characteristic values?

One solution is what might be called first-class modes:

(element FOO
  (let ((x (compute-some-info-about-foo))
        (y (compute-some-other-info-about-foo))
    (with-first-class-mode
      (first-class-mode ; new syntax, note that the values of x and y in
effect get passed down
        (element BAR
           (if x ; NB
              (make simple-page-sequence)
              (make scroll)))
        (element BAZ
           (literal y)))
      (process-children))))

This is quite similar I think to the sort of thing Matthew Fuchs as talking
about in his critique of the DSSSL transformation language at SGML 96.

James







 DSSSList info and archive:  http://www.mulberrytech.com/dsssl/dssslist


Current Thread