UA requirements vs. document conformance (Was: Re: Choosing name for XML serialization)

On Jun 25, 2007, at 10:28, Julian Reschke wrote:

> Ian Hickson wrote:
>> On Mon, 25 Jun 2007, Julian Reschke wrote:
>>> Ahem. How can you say it's compatible if - for instance - a  
>>> widely used feature like the style attribute is removed?
>>> What am I missing here?
>> There's a distinction between what is conforming (what authors are  
>> allowed to do) and what is supported (what browsers are required  
>> to do). Just because authors are (say) not allowed to use the  
>> align="" attribute on the <p> element, it doesn't mean that if  
>> they _did_ use the align="" attribute, that it would not work. And  
>> indeed, in the "rendering" section of the HTML5 specification we  
>> will be describing all manner of things, possibly including things  
>> like <marquee>, align="", etc. That section is still not done.
> Ok, so browsers will still be compatible, but the documents will  
> not be conforming anymore. To me, that seems like an extremely bad  
> idea, potentially driving people away from checking document  
> conformance. On purpose?

That browsers need to support legacy features is not up for debate.  
However, what is deemed conforming is.

If we allow all the legacy stuff as conforming, one set of people  
will think we are nuts. If we don't allow something as conforming,  
another set of people will think we are nuts for making something  
that browsers support non-conforming. (A set of people will think  
that we're nuts for specifying browser behavior for legacy stuff, but  
that's something we need to do in order to write an honest spec for  

The notion of document conformance is partly a way to try to sway  
author behavior to a "good" direction, partly about helping authors  
make something sensible and partly a way to help authors avoid  
shooting themselves in the foot with something that doesn't work  
reliably across legacy UAs.

Deciding how far to take the swaying towards "good" part is a  
balancing act. Allowing everything that "works" in browsers will  
water down the goal of shunning some "bad" authoring practices and  
the goal of helping authors make something sensible. On the other  
hand, making the notion of document conformance too strict so that it  
doesn't meet the wishes of most authors will make authors ignore  
conformance as something that causes too much trouble for no  
understandable benefit.

Moreover, sometimes allowing something "bad" is less bad than  
forbidding it. For example, banning target='' would cause badge  
hunters to open windows using JavaScript which is harder to filter  
out if the user doesn't want pages opening windows.

I think style='' is mostly "bad" but we should allow it on all HTML  

Henri Sivonen

Received on Monday, 25 June 2007 07:51:02 UTC