Re: Associating DSSSL style sheets with documents

Eve L. Maler (elm@arbortext.com)
Fri, 14 Mar 1997 17:34:50 -0500

At 02:53 PM 3/14/97 -0600, Len Bullard wrote:
>Eve L. Maler wrote:
>>
>>
>> The difference is that, by convention, you're making PI markup available
>> that's available to every document and to every *location* in a document if
>> necessary, no matter what its DTD (and no matter whether it even has one).
>> It just happens to look suspiciously like a start-tag, which may be helpful
>> to any software that has to parse the PI string.
>
>By convention? You mean, by application.

I'm not sure I catch your distinction. If we agree on a meaning and a
syntax for it, we've made a convention. (Like when everyone asks "How are
you?" and expects a short, positive answer. :-) Applications can now
predictably act on the usage of the convention. (Like when someone starts
to walk away after a moment, safe -- usually! -- in the assumption that the
other person just answered "I'm fine.")

>An inclusion on root makes an empty element available
>to every location. A PI is something every document has to have.
>That isn't an improvement. If you use a DOCTYPE and know the DTD,
>don't
>you get the same effect? XML goes out it's way to load up an
>instance just to get around a DTD. I question the utility of that.
>We tell them they are being freed of fixed markup, then add a
>question mark and say, oh, that's OK, that's XML.

But XML doesn't have inclusions, and any one document may not even have
DTDs. So your "ifs" sometimes don't come true. I agree that we don't want
to push legitimate DTD functions into PIs, which give you a lot less
validation power. But processing instructions (in the regular English
sense) don't belong in the normal markup scheme most of the time.

>> I don't think links in general should be done this way, but I do believe in
>> PIs being used for, uh, instructions to processors.
>
>Ummm... sure. Sort of what links are.

Well, a reference to a stylesheet is surely a link, but not all links are
references to stylesheets. Also, not all processing instructions are links
to something. Do you think PIs are never appropriate?

>> (In other words, I'm
>> not 100% against PIs, as some people are.) In particular, I'm starting to
>> get very fond of PIs for anything that has to be specified per entity.
>
>No doubt.
>
>len

Eve