- From: Paul Prescod <papresco@calum.csclub.uwaterloo.ca>
- Date: Wed, 23 Apr 1997 20:49:13 -0400
- To: Todd Fahrner <fahrner@pobox.com>
- CC: www-style@w3.org, Jukka.Korpela@hut.fi
Thanks for your interesting note Todd. I've been struggling with this cascading issue myself recently. I support the cascade's goals, but every time I think about adding them to DSSSL (just as a thought experiment!) I think of how messy any system that uses them would be, and how seldom they would actually be useful. It is too general a solution to a very specific set of problems. I feel that these criticisms apply to CSS as much as to DSSSL. The Dangers of Combination ========================== Combining three levels of styles (UA, author, reader) without communication between the parties is probably not a good idea as you point out. It doesn't result in good style. It doesn't result in good accessibility. And as a "computer person" it bothers me because it can surpress authorial meaning without warning. These issues are well explored at the URL Todd gave: http://www.hut.fi/~jkorpela/styles/harmful.html Unfortunately, he doesn't really address the issue of generic markup. Documents will always need different structures. Whether it is achieved through CLASSes or new tags, it is necessary and it is coming, and we must deal with it. I'm not ready to receive arbitrary SGML documents with no stylesheet and I presume that nobody else is either. The Uselessness of Local Override ================================= Any interesting override probably should not be applied on an element by element basis. If you are colourblind, then you need a way to remap ALL colours, used ANYWHERE, even on elements and classes you have never heard of, to black and white. If you are nearsighted, you need to remap all small fonts. This suggests to me that CSS is actually a step backwards. In my UA today (IE 3.0) I can already make all of the fonts bigger with a button click (my wife uses this). I certainly hope that in the future I don't have to write a stylesheet for it (especially one per DTD!!). Let's face it, if you have special needs visually you will simply turn stylesheets off or use Lynx. Thankfully most HTML documents are still visible this way. CSS has played a major part in saving HTML from becoming a huge presentational mess that would not be useful from Lynx. So CSS is already a huge success in my book, but it is the SS part, not the C part which has saved HTML. As far as I know cascading is hardly ever used except as a short cut for "import the default UA stylesheet with these overrides." So CSS is a success but we could drop the cascading and it would continue to be. I advocate that. Generic Markup ============== What about XML? Sadly, in some ways XML will be a step backwards for the visually impaired. That won't be able to predict tag names and ignore stylesheets the way they do with HTML+CSS. On the other hand, the information they are getting will be much more rich and perhaps navigable in very powerful ways that can be optimized for their special needs. But cascading is useless in an environment with no fixed tag names. They need plain old override and they need a convenient user interface for doing so. I think that they also need parameterized stylesheets, which is somewhere that DSSSL will shine. We can allow them to "pass in" a font-size and choose all others as percentages of that. They can "pass in" whether they need black or white display and the stylesheet author can predict those needs. Instead of the general facility available in CSS we must start asking people: "What is your disability" "What do you need changed" "How can we change it." In other words, if the web is going to be more diverse in the future, we must give responsibility back to the authors to do the right thing, not require the users to become masters of overriding stylesheets for hundreds of different DTDs (or class-based variants of HTML). Where a user is lucky enough to stumble upon a DTD that they are familiar with (HTML, DocBook, TEI), they should a stylesheets, made, perhaps, by W3C or the Yuri Rubinsky foundation, tailored for them. There is no good reason to cascade the user's stylesheet with the author's if the user already has a complete one. We should also encourage authors to stick to well-known DTDs unless there is a good reason to go beyond them. The war for valid documents isn't over! It's just about to begin: XML should be used to give people freedom, but also to re-enslave them to well-defined structures. Summary ======= In short, I think that the cascade is a technological solution to a problem that cannot be solved technologically (unfortunately). 3rd parties must make accessible stylesheets. Authors must take responsibility for accessability. Casual users must stick to well-known DTDs (or DTD variants as in HTML w/ a particular CLASS set) with an existing accessibility infrastructure. Professionals and corporate users should only undertake to create a new DTD, tagset or "class set" when they intend to back it up with an accessibility infrastructure. What I mean by accessibility infrastructure is stylesheets, parameters, public identifiers, "proper use" documentation, etc. We have a lot of work ahead of us. I don't expect that cascading will play a big part in that equation. Paul Prescod
Received on Wednesday, 23 April 1997 21:33:32 UTC