- From: Paul Prescod <papresco@calum.csclub.uwaterloo.ca>
- Date: Thu, 20 Jul 1995 08:52:11 -0400
- To: www-html@www10.w3.org
At 12:50 AM 7/20/95 -0400, Ka-Ping Yee wrote: >To reference files on other servers would not be impossible -- it's just that >if we're going to do it, we might as well properly hash out a way to >reference it in HTML instead of doing hacks to be parsed by http daemons. The important thing to recognize is that SGML has already worked out all of the symantics for "including" text from a document fragment into another document. SGML External entities are well defined and well document. At first glance it seems that the simple answer would be: <!ENTITY foo SYSTEM "http://www.foo.com/fragment.html" > ... &foo; Of course this would bring the browser vendors one step closer to using real SGML-based parsers. This goes against what seems to me to be one of the unspoken mandates of the HTML WG: to allow browser vendors to get away with writing ad-hoc parsers, and to force SGML tools to conform to the resultant chaos. >The problem with including arbitrary bits of HTML is that you have no >guarantee other documents will be conforming documents. In fact, it would >not be possible to directly include another conforming document (with >another <head> and <body>) within a conforming document. Even two >fragments of correct HTML can cause havoc when one is included within >the other (to wit, imagine nested <A>s or <FIG>s). How is this a problem? How is it any different if I do a server-include or CPP or write a Perl script that combines the documents? The only difference is that someone else may control the linked data. Whenever you link to something elsewhere you run the risk of the linked data changing. But if someone changes the markup at their site and it renders my document invalid, I will be at worst embarrased. A much worse situation can occur with a simple IMG. If you link to a picture, the maintainer can change it to something illegal or obscene. If you choose to include someone elses text or image in your home page, incorrect markup is the _least_ of your worries. =) HTML 3.0 handles this with the MD attribute and perhaps the entity reference syntax should be extended to also allow a checksum. >This is why i believe included documents should be treated in a similar >fashion as figures: they are set apart in their own box (that is, >logically speaking -- the box doesn't have to be physically visible) >and continue to be treated as separate documents. You are suggesting a mechanism for implementing full subdocuments. I am suggesting a mechanism for implementing document fragments. Both are necessary. Paul Prescod ---------------------------------------------------------- HTML Myths Page: http://www.incontext.ca/~papresco/htmlmyth
Received on Thursday, 20 July 1995 08:52:32 UTC