- From: len bullard <cbullard@hiwaay.net>
- Date: Fri, 14 Feb 1997 09:41:34 -0600
- To: Murray Altheim <murray@spyglass.com>
- CC: w3c-sgml-wg@w3.org
Murray Altheim wrote: > > bosak@atlantic-83.Eng.Sun.COM (Jon Bosak) writes: > >[Murray Altheim:] > > > >| The two methods described perorm the same function, but are used for > >| different purposes. The first example allows a designer to create an > >| new SGML application. The second allows legacy applications like > >| DocBook or HTML to be retrofitted with architectural form attributes > >| to be used within an XML processing system. > > > >Now I'm hearing arguments in favor of using the attribute method. I'm > >not hearing anything about why we need both methods. Either the > >attribute method works or it doesn't; if it does, why do we need both? > > Well, I'm certainly an advocate for attributes, but I can imagine a family > of SGML applications that conform to XML, ok, a family of XML applications. > Within each, the hypertext mechanism could certainly be handled by AF > attribute, but my little imagination is thinking of users whose > understanding is that the base functionality of an element comes by way of > its GI, not via its attributes. It's a bit of a stretch for some people to > think you could transform > > <B>important text</B> > > to hypertext by adding something akin to > > <B XML-LINK="alink" HREF="http://www.foo.com">important text</B> > > It's an appealing idea (on cursory examination at least), but probably > confusing to the unwashed masses(tm). > > We certainly don't want AF understanding to be a requirement for XML use, > and while a simple explanation might suffice, it seems that simply stating > that *one* method of providing hypertext markup would be to use the > <XML-LINKTO> or <XML-ASSOCIATE> elements (or whatever they end up being > named). It's logical, but my experience is the great unwashed masses don't design DTDs or use arch forms. That argument is overrated. If I have to live with one and only one way of doing it, I prefer to have the GI left to my discretion, and the superclass assignment in the attribute as a silent but helpful partner. After all this talk of extensibility and something more powerful than HTML, I hate to see us return to precisely the same design by giving them a non-extensible set of link types. They don't get the power without the responsbility to learn the technique. len
Received on Friday, 14 February 1997 10:41:35 UTC