[whatwg] several messages about XML syntax and HTML5

[Ian Hickson]

On Mon, 4 Dec 2006, Sander Tekelenburg wrote:
> >
> > Second, it isn't the browser's best guess. HTML5 defines error 
> > handling in detail, so it doesn't matter if the page is conformant or 
> > not, it's still going to be interoperably handled.
> 
> Surely you're not saying that HTML5 will define error handling for every
> possible case a UA may run into?

Yes. In fact, not only will it define this, it already _does_ define this. 
e.g.:

   http://www.whatwg.org/specs/web-apps/current-work/#parsing

...covers the parsing of HTML into a DOM, and describes how you handle 
every single error, including error recovery.


> No doubt UAs will run into invalid HTML5 documents, and apply best 
> guesses -- because the user won't accept a "this browser refuses to 
> render invalid documents" message.

The requirement is not that the browser refuse to render the invalid 
document. The requirements vary from error to error but are always precise 
ways for how to handle errors gracefully.


> That aside, HTML5 UAs will continue to have to best-guess all those 
> billions of existing invalid pre-HTML5 documents.

The specification was very, very carefully written to take into account 
those pages, so that an implementation of the spec handles them in a 
manner compatible with legacy UAs, without having to do any guessing. (I 
did studies of several billion documents to support this work.)


> >> But then you simply implement it so that you get a happy smiley when 
> >> the page is valid, and nothing at all when it isn't.
> >
> > This would still fail usability testing, though for different reasons. 
> > Now the reason is that the user would think the browser was broken, 
> > because it randomly, on about 5% of pages, would show a smile.
> 
> How is that different from that key in the status bar that they see 
> randomly on some pages?

Given the extremely bad usability of SSL UI, and the fact that the 
security community is currently having to desperately find new ways to 
make sites secure in a way that users understand, your analogy is actually 
very apt. There are a number of studies that show that SSL UI is 
horrendous; one study I read suggests that over 60% of users don't even 
pay attention to SSL error messages, let alone the lock.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'