W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > January to March 2000

RE: A "one size fits all" personalized web page?

From: Charles McCathieNevile <charles@w3.org>
Date: Sat, 1 Jan 2000 10:22:24 -0500 (EST)
To: Scott Luebking <phoenixl@netcom.com>
cc: A.Flavell@physics.gla.ac.uk, w3c-wai-gl@w3.org
Message-ID: <Pine.LNX.4.20.0001011004330.27823-100000@tux.w3.org>

my point about semantics in HTNML and my point about poor design being
allowed to persist seem to me closely related. Of course there is no
information about sports in the example you have given, but that is not an
artifact of it being coded in HTML, rather it is that you have not supplied
that information in the first place.

If you look through the WAI Authoring Tool Guidelines you will find code such

  <a href="#def-something" rel="glossary">something</a>

which uses HTML coding to provide semantic information. The move towards XML
and RDF are to make it easier to customise this information, but it is
perfectly feasible and reasonably common to do it in native HTML, although
there is a serious problem with tool developers being backward about creting
tools that support this. You should look through the HTML specifcation for
the attributes title, rel, and class in particular, and for elements such as
blockquote, samp, cite (also an attribute) which are designed to code
specific types of semantics (the class mechanism allows for the general
encoding of semantics).

Some of the features of poor design that I am arguing against are:

departing from the notion that everything that is useful should be a
first-class object - in short, that it is harmful if the information that one
link is related to sprot and another is not is only available to a back-end
engine which will release it or not based on its interpretation of user
needs. This is a fairly central part of the way I understand Tim's design for
the web, although it is a principle applied much more widely in general
information systems design. This is also the major problem I have with the
approach you have described - it takes away from the user the ability to know
what is going to be given to them, by doing the interpreting of what they
want for them, based on further information that is not available to them. I
think this is also linked to the discomfort people expresed about getting a
"different" version of content. Of course it could just be that I am
misunderstanding the way your material is built in practise.

promoting the continued existence of sites that do not meet user needs, on
the basis that an alternative is available. This is a repair strategy, not an
aim. I believe that Chuck and I are expressing similar ideas, in that
Universal Design is the goa we would like to achieve, and feel it is
possible. At the same time, we are explicitly reognising that there is a
danger in assuming that one answer will fit everybody - although this is the
principle behind universal design it is only true if the desin is good, and
that always needs to be tested more. However, the risk is not significantly
altered if there are a couple of alternatives - as I pointed out with the
scenarios I wrote earlier in this thread, there are a large number of
alternative cases to be covered (which is why I think Universal Design
principles offer the most effective and efficient approach to solvin the



On Wed, 29 Dec 1999, Scott Luebking wrote:

  Hi, Charles
  Read through the pages on the Semantic Web.  Why is the Semantic Web needed?
  I believe you might be confusing strings of text with semantic information.
  They are not the same.
  For example, suppose that a web page is about Chicago.  Let's say that among
  links there are:
      <A src="#chicago_cows.html"> Chicago Cows </A>
      <A src="#chicago_bulls.html"> Chicago Bulls </A>
  The strings of text are "Chicago Cows" and "Chicago Bulls".  Now, suppose that
  the web page user wants the links related to sports to be at the top of
  the page.  If a user wanted his user agent to put the links related to
  sports at the beginning of the page, how would the user agent know whether
  the "Chicago Bulls" link or the "Chicago Cows" links are related to sports?
  The semantic information is lost when the HTML is created.
  For a personalized web page, information about the category of each link
  can be stored in a database.  The links can be ordered according to
  whether the link is sports related or not and then the HTML is created.
  Are you objecting to personalized web pages?  Are you saying that it is a poor
  solution for meeting user needs?
  What is your definition of "badly designed"?  I believe a beautifully
  graphic personalized web page can be a good design for a large number of
  users.  Why should a personalized web page have to accomodate all
  users?  The more important issue is that each user gets the information
  he wants in the format most easy for him to use.  I think that what
  Chuck posted from the CAST web site really makes a lot of sense.
Received on Saturday, 1 January 2000 10:24:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:47:01 GMT