Re: mixed signals on "Writing HTML documents", tutorial, etc.

Ian Hickson wrote:
 > On Tue, 19 Jun 2007, Philip & Le Khanh wrote:
 >> How /can/ HTML 6 be "compatible" in the sense that a validator would
 >> understand the term? Either a document is HTML 5 or it is HTML 6
 >> (or it is something else); it cannot be both HTML 5 /and/ HTML 6 unless
 >> HTML 6 neither adds to, nor eliminates from, HTML 5, in which case
 >> there would be little if any reason to introduce it.
 >
 > Why not?
 >
 > The following:
 >
 >    <!DOCTYPE HTML>
 >    <html>
 >     <head>
 >      <title> Demo </title>
 >     </head>
 >     <body>
 >      <p> Hello! </p>
 >     </body>
 >    </html>
 >
 > ...is, modulo the DOCTYPE line, conformant to all versions of HTML ever
 > made, as far as I know, going back at least to HTML2. Why would it not
 > also be conformant to HTML5, HTML6, HTML7?

Yes, the question was poorly put.  Of course it is possible
to construct a trivial example that (modulo the DOCTYPE
directive) is conformant to at least two versions of HTML,
if not more.  The question should really have been "How
can all valid instances of HTML 5 also be valid instances
of HTML 6 and /vice versa/, if HTML 6 neither adds to nor
removes from HTML 5 ?"

Because this is the question that is at the core of things.
A validator is not required to validate one trivial example :
it is required to validate /all/ examples with which it
is presented, and thus there are two possible scenarios :
(1) It can determine unambiguously from the DOCTYPE
with which version of HTML the document claims to be
conformant, and then it can verify that claim; or (2)
it cannot tell from the DOCTYPE with which version of HTML
the document claims to be conformant, in which case it
has to validate the document against all possible
specifications that require nothing more explicit than
<!DOCTYPE HTML> and report with which of these, if any,
the document conforms.  What it cannot do is assume
(and behave as if) that all valid instances of HTML 5
are also valid instances of HTML 6 and /vice versa/,
unless HTML 6 is isomorphic [1] to HTML 5, in which case
there will be no point in adopting the former since
everything that can be expressed therein can be
expressed identically in HTML 5.

 > The merits _aren't_ obvious. The drawbacks are many, and serious, but the
 > benefits seem slim and dubious. (This has been discussed to great lengths
 > in this mailing list over the past months.)

The benefits seems so obvious (and so essential), and
the drawbacks so minimal,  that I am at a complete loss to
understand how the proposal to drop versioning can
ever have arisen.

Philip Taylor
--------
[1] characterized by a one-to-one correspondence between the
     members of two or more sets

Received on Tuesday, 19 June 2007 15:35:09 UTC