Re: The argument for |bugmode|

On Apr 20, 2007, at 00:03, Chris Wilson wrote:

> No.  We will have our own proprietary, non-invalidating opt-in  
> switch to "really standards as of IEn" mode.

With my conformance checker developer hat on, if there's no way to  
convince MS not to have such a switch, I'd much rather work with you  
and the WG on agreeing on an attribute for the purpose of *document*  
conformance instead of seeing you come up with a switch that is baked  
into syntactic features that are not supposed to carry significance.  
(However, I'd also like to define *UA* conformance in such a way that  
other browsers are free to ignore the IE opt-in switch.)

A switch that is cloaked from conformance checkers is also cloaked  
from sane XML tools. (And people will use XML toolchains with text/ 
html as well be prepending an HTML5 parser to their processing  
pipeline.) My experience from working for a vendor whose product  
round-tripped content using XML tooling is that round-tripping  
syntactic features that were not supposed to have significance is  
expensive and bad for software architecture.

I don't like having an IE opt-in switch, but if you are going to make  
sure that the net effect is that the switch isn't flagged as non- 
conforming no matter what, I think it is better to do it in a way  
that minimizes harm to intermediate tool vendors who (rightly or  
wrongly) will believe that it is their duty to round-trip the opt-in  
switch.

My order of preference from better to worse is this:
  * No opt-in switch at all.
  * Opt-in switch in an attribute on the root element.
  * Opt-in switch in a comment.
  * Opt-in switch in a magic system id of the doctype.
  * Opt-in switch baked into the choice of character data escaping,  
choice of attribute quoting or choice of amount of whitespace between  
attributes.

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