Re: An HTML language specification

Andrew Fedoniouk wrote:
>
> Henri Sivonen wrote:
>>
>> On Nov 14, 2008, at 20:42, Mark Baker wrote:
>>
>>> I think we've had this discussion before 8-)
>>>
>>> As often happens, Roy says it better than I could;
>>>
>>> http://lists.w3.org/Archives/Public/public-html/2007Nov/0430.html
>>
>>
>> It seems to me that Web authors seek to elicit browser behaviors when 
>> they write HTML/CSS/JS. They don't just put abstract meaning out 
>> there. (In the common case; I'm sure you can show an existence proof 
>> of the opposite.) There are other kinds of Web clients, too, but 
>> generally people publish content primarily to enable people to access 
>> the content using browsers and the ability of other programs to 
>> consume the content is just a bonus. (Of course, there are 
>> exceptions: Certain types of black hat SEO target the behavior of 
>> search engines and the content doesn't need to be useful in a browser.)
>>
>> I don't think it's useful to try to decouple "the language" from 
>> "browser behavior" normatively. However, I think that producers of 
>> HTML documents would benefit from an informative document that gives 
>> instruction on how to produce conforming HTML5 documents. That is, 
>> the following the guidance of the informative document should result 
>> in conforming documents but the guidance wouldn't need to give all 
>> the possible ways in which a document could be conforming.
>>
>> I think we don't need a separate normative language spec but instead 
>> we need an informative authoring guide ("primer" in the W3C lingo, 
>> although that word isn't great in the title of a document aimed at 
>> global audiences). We already even have an editor (Lachy) signed up 
>> for the authoring guide.
>>
>
> For me HTML5 specification appears as an attempt to provide unified 
> document targeted on implementors (of HTML5 parsers) and content authors.
>
> But...
>
> As an example consider these two sub-chapters:
>
> 4.4.10.1 Creating an outline
> 4.4.10.2 Distinguishing site-wide headings from page headings
>
> of the chapter: http://www.w3.org/html/wg/html5/#headings-and-sections
>
> As an implementor of HTML5 parser I do not understand what to do with 
> information provided there. It appears as it is useless for the need 
> of parsing input stream into the DOM.

As an implementor of a /parser/ it is not significant. As the 
implementor of a browser it may well be significant e.g. if you want to 
provide an outline navigation mode. As an author it is significant 
because you need to understand how UAs determine document structure in 
order that you can translate the document structure that you want to use 
into html. Of course the vast majority of users will get this 
information from a tutorial rather than from the spec but it is 
important that the people writing the tutorials have access to complete 
and correct information.

> Question Générale: Whom this document is aimed for?

The HTML 5 specification? Implementors and "language lawyers" i.e. 
people who need a detailed and authoritative answer to questions about 
HTML 5 and its processing in UAs. Expecting these people to understand 
concepts such as the DOM is not really asking too much.

I am wary of any attempt to create a second, normative, document aimed 
only at authors. Every time there is more than one "definitive" source 
for the same information there is the possibility of conflicts, leading 
to confusion. However, as the example above shows, there is much that is 
of importance both to implementors and to authors in the spec and 
performing a full separation is impractical. Moreover any normative 
language spec would, on account of the need for normative levels of 
precision, still only appeal to language lawyer types. The set of people 
interested in such an inapproachable document but uninterested in any 
aspect of UA behavior seems to me to be small (although I suggest that 
many such people would be involved with the W3C, so we might have a 
selection bias affecting our priorities here). Ordinary authors are 
considerably more likely to be interested in a human-friendly, 
informative, authoring guide with links to the appropriate normative 
text for cases where precision and authority are needed. I believe such 
a guide is already being worked on.

Received on Sunday, 16 November 2008 09:48:19 UTC