Re: Possible use of <INSERT> - serious inquiry (fwd)

MegaZone (megazone@livingston.com)
Tue, 19 Mar 1996 20:30:44 -0800 (PST)


From: MegaZone <megazone@livingston.com>
Message-Id: <199603200430.UAA22112@server.livingston.com>
Subject: Re: Possible use of <INSERT> - serious inquiry (fwd)
To: www-html@w3.org
Date: Tue, 19 Mar 1996 20:30:44 -0800 (PST)

Once upon a time Murray Altheim shaped the electrons to say...
>A valid HTML document is of the form:
>
>    <!DOCTYPE ...>
>    <HTML>
>    <TITLE>...</TITLE>
>    <BODY>
>    ...
>    </BODY>
>    </HTML>
>
>To insert another valid HTML document anywhere in this structure
>invalidates the document, unless the inserted entity is a document fragment
>and it is allowed at the inserted location. I don't think I've ever seen
>any proposed resolution to this problem, where in practice folks will
>probably simply insert complete HTML documents. It would be a royal mess,
>as it puts a heavy requirement on authors to do their own entity
>management, something I've seldom seen in practice.

Either the browser has to handle this and ignore everything *except* the
contents of the <BODY> - which is easily done - or the user has to use
fragments with no wrapping tags.

I prefer the first solution because the documents could then be validated.

This is no different from the ability *NOW* to use Javascript to generate
HTML on the fly as the page is loading, and you could just as easily
generate BAD HTML this way.  Same with CGI generated content and JAVA.
So I don't see it as a new problem.  Actually, since this is read into
the browser as it loads, I see this as easier to defend against.

>Also, the issue of whether or not the inserted document exists on the same
>system may or may not be a problem, depending on your SGML system. While a

Well, if that is a problem it is already a problem with the current
<INSERT> tag.  How does the validator know what is being included?  It
might be something that generates bad HTML once loaded.

>validating parser may be able to locate external entities via URL, this is
>not a requirement in SGML (ISO 8879:1986), it is simply a nice feature of
>nsgmls. Dan Connolly and I have gone 'round on this one before.

Well, IMHO that's just too bad for the validator engines that aren't
bright enough to either ignore this or follow it.  Why should everyone
have to limit themselves because of something like that?

>Therefore, any feature that required an insertion of content based upon
>URLs may not be conformant. Someone may correct me on this, but this is the

I'm basically in the "I don't really care" section on the "Is HTML to be
SGML compiant or its own language" war.  If a feature is useful and in
demand, but it happens to break SGML - so be it.

If my pages are valid HTML but not valid SGML - frankly I don't care.

-MZ
--
Although I work for Livingston Enterprises Technical Support, I alone am
responsible for everything contained herein.  So don't waste my managers'
time bitching to them if you don't like something I've said.  Flame me.
Phone: 800-458-9966  support@livingston.com  <http://www.livingston.com/> 
FAX: 510-426-8951    6920 Koll Center Parkway #220, Pleasanton, CA 94566