[racket] comments requested from users of SXML and PLT xexprs

6 minutes ago, Neil Van Dyke wrote:
> Thanks for your comments, Eli.  I am still wavering on whether to do
> the permissive thing.  And I am also looking at whether Typed Scheme
> and/or "html-template" make the permissiveness less useful than
> before.

Well, the problems that I ran into wouldn't have been different if
there was a type system around -- changing a function still means
changing call sites, etc.

[A little unrelated to this, I'm wonder whether you could use TR to
improve the error situation but (a) I suspect that it would be similar
to making a new `splice' struct that couldn't be added to an xexpr
without explicit conversion (and splicing), and (b) if it can be used,
then I suspect that the type errors that are involved would be hard to
decipher.]


> Permissiveness sure does make some things harder to implement and
> more computationally expensive.

Why is it harder?  My experience so far is that the cost in terms of
implementation complexity is very minor.


> I was speaking of named and numeric character entity references in
> attribute values, which are supported in some versions of XML and
> HTML.  "Composition" was a poor choice of term on the Racket email
> list.
> 
> I'm not sure what you mean by "combining `style' attributes".  Is
> that a transform rather than a representation/linguistic issue?

It's a different problem.  Imagine that you define `bp' as (using
`scribble/html' terms):

  (define (bp . body) (apply p style: "text-weight: bold;" body))

it works, but I'd like to be able to use it with:

  (bp style: "color: blue;" "blah blah")


too.  (And you can see how that goes...  I could make `bp' pull out a
`style:' value and string-append it to the weight, but...)

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                    http://barzilay.org/                   Maze is Life!