detecting unmarked navbars

Every now and then I wonder about all the markup that we invent.

I am not going to suggest that we retract the CLASS="nav" 
proposal.

But I want to talk for a minute about how the user agent could
figure this out in the absense of author markup.  The
distinguishing characteristic of a navbar is that it contains
links and little else.

What kind of a pattern would be very likely to match navbars
and little else?

We can hypothesize an overall form or pattern for a navbar:

* at most one navbar framing and introduction

* for each href a small surrounding structure which fits a
  consistent link-presentation pattern

* optionally, inter-link punctuation that is small and 
  is consistent [c.f. <space><vertical bar><space>]

One can inductively construct a rationale for containers in a
document explaining how that container matches this pattern.  The
link-presentation pattern is a subtree pattern constructed per
the DOM schema which contains any relatives of the href attribute
for which a comparable relative exists for all hrefs in the
container that we are working on.  The inter-link punctuation
form is not a structural pattern, but a literal hypertext
fragment which appears in the interstices between instances of
the link-presentation pattern.  The largest literal fragment that
appears in all such inter-link positions is recognized as the
hypothetical inter-link puctuation module.  Etc.

Then rank the containers in terms of how simple the rationale is
and how well it fits.  One can set thresholds that consider
ordinal ranking (most likely navbar substructures in this page)
and absolute grading (most like the generic-navbar hypothesis,
regardless of context or peers) in order to decide which of the
substructures in the page to default to being treated as navbars.

Al

Received on Saturday, 14 November 1998 13:14:14 UTC