W3C home > Mailing lists > Public > www-qa@w3.org > September 2001

Re: Quality Tips suggestion: Use <h1> for titles

From: Al Gilman <asgilman@iamdigex.net>
Date: Mon, 17 Sep 2001 14:44:56 -0400
Message-Id: <Version.32.20010917082610.0409bf00@pop.iamdigex.net>
To: Alex Rousskov <rousskov@measurement-factory.com>, Aaron Swartz <aswartz@upclink.com>
Cc: www-qa@w3.org, wai-tech-comments@w3.org
If each of these answers is bad, in some way, what is the question?  What is
the space in which we are looking for a good answer?  What are the
considerations that contribute to an answer being good or bad?

Does this have something to do with automatic recognition of an infrastructure
for navigation?


In my rough sense of the bidding, the following propositions might enjoy
varying universality of goodness:

+ [known relation] The scope of what a heading heads should be unambiguous.

+ [standalone] A heading, taken by itself, should provide a good discovery
representation for what it heads.


+ [standalone minus one] A heading, taken in the context of its immediate
super-heading, if applicable, should provide a good discovery representation
for what it heads.

+ [nested] The scopes headed by different headings should be non-overlapping.

+ [monotone] If the scope of header A is completely within the scope of header
B then the header level index of header A should be a higher ordinal than the
scope of header B.  [strict monotonicity of index in order of _depth_]

+ [1-justified] The indices used in header types should be consecutive and
start with 1.

At the moment I doubt that a realistic media analysis of the graphical
of web design would support the notion that the 'nested' proposition is a
universal principle.  If this is the case, then the '1-justified' principle
just be debating angel populations on pinheads.

For intrapage navigation [see Jon's list of UAAG lessons learned as input to
XHTML 2.0 requirements cited above] we need a block structure that is
and represents a conceptual structure  for the content readily grasped by the
user and readily endorsed by the author.  Well-ordered nesting is negotiable. 
When it's there, it's easier to understand the shape of the space you are
navigating without visual context cues.  But it's harder to map the actual
structure of the topic and hence the screen-rhetorical exposition of the topic
into such a strict form than into a general Venn diagram of topics.  See
backups to 


for the most conspicuous example that enjoys formal representation in current
HTML, i.e. table columns and table rows intersect, they don't nest either one
in the other.

Sorting this one out will lead us straight to Booch.  That is to say that we
need to understand the role of generic/specific relationships and not just the
part/whole abstracted from the syntax to capture how the topicality of the
syntactic units relate.  In a system of scopes that is in a form simple
for the user to remember as the navigate the page and rich enough to represent
what the author has to say, some tree levels may be non-notional and some of
the notional scopes may overlap and hence come from outside the pure tree.


At 01:28 PM 2001-09-16 , Alex Rousskov wrote:
>On Sun, 16 Sep 2001, Aaron Swartz wrote:
>> On Sunday, September 16, 2001, at 11:37  AM, Alex Rousskov wrote:
>> > Often, webmasters have to use <h2> for first-level titles.
>> Why do you "have to"? Is it an absolute necessity that all your 
>> users see the title smaller? Does your page not use <font> tags? 
>> I find it hard to believe that this would ever be needed.
>Using <h2> is not absolute necessity. Using <h1> is not either. This
>is not a question of absolute necessity. This is a question of what a
>person writing/generating HTML should use. Both <h1> and <h2> are bad
>(for different reasons discussed on this thread).
>If I write/generate relatively simple pages that do not need CSS to
>look good, I will not introduce CSS just to make picky HTML parsers
>If I write/generate pages that have to rely on CSS for other reasons,
>your suggestion is perfectly valid.
>My point is that CSS is not acceptable as a "general recommended
>workaround". Otherwise, you should just recommend to always use XML +
>CSS (or similar) instead of HTML.
Received on Monday, 17 September 2001 14:42:48 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:40:28 UTC