Re: Validation error frequencies

On Feb 6, 2008, at 15:44, Ben 'Cerbera' Millard wrote:

> HTML5 validators need not be blunt instruments to hit people round  
> the head
> with until they put every byte where we want it. HTML5 validators  
> can be
> friendly and helpful things to let people know when they do unreliable
> things and provide pointers to the reliable ways.

Right. I want HTML5 conformance to be formulated in such a way that it  
makes sense when Web authors are taken to be the primary users of  
validators. It's a bad idea to make products that don't work for their  
primary users.

> If a thing is reliable, the validator should not pester people about  
> it,
> imho. The authoring chapter/document of HTML5 can give advice about  
> which of
> the reliable things are best suited to specific situations.

I agree. To put things into perspective, the type attribute is just  
cruft in <style type='text/css'> and in <script type='text/ 
javascript'>. Yet, the draft allows both. <script  
language='JavaScript'> is no more crufty than those cases, but it is  
disallowed.

And unlike type='text/javascript' or type='text/css', border='0'  
actually does something useful (albeit not in the most efficient way  
possible).

[...]
> HTML4 Strict makes absent many presentational features, such as <img  
> border>
> and <font face>, in favour of style sheets. I guess the presence of
> cellspacing and cellpadding are bugs.

Back when HTML 4.0 proceeded to REC, CSS2 was not a REC, yet. CSS1  
didn't cover tables, so presentational attributes on tables were  
needed in practice and, thus, they stayed in HTML 4.

Now, a decade later, the CSS alternative for cellspacing still doesn't  
work in the browser with the most market share, so telling authors to  
just use CSS still isn't practical.

Considering what needs to be in or out of HTML5 in the light of HTML  
4.01 Strict misses two things:
  1) HTML 4.01 Strict didn't deprecate presentational attribute  
categorically (so it isn't precedent for categorical deprecating/ 
obsoleting presentational stuff): keeping <table cellspacing> but not  
<img border> doesn't mean that they are fundamentally different--just  
that their substitutability with CSS was different in 1997.
  2) the status quo from which authors will be moving to HTML5 is  
Transitional--not Strict.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Monday, 11 February 2008 18:03:29 UTC