[whatwg] After attribute name state

[Ian Hickson]

On Sat, 7 Jan 2006, Simon Pieters wrote:
> 
> The After attribute name state section[1] now says that the attribute 
> value will be the empty string for "Anything else", which is not true in 
> all cases. (I know it's WIP, and I know you know about this, but still.)
> 
> Boolean attributes. <table border> is the same as <table border="1">, 
> and contenteditable is the same as contenteditable="true" (at least in 
> IE, not in Opera). There are probably some other exceptions aswell.
> 
> [1] http://whatwg.org/specs/web-apps/current-work/#after

Yeah... I'm still trying to work out how to handle these cases. I need to 
compare cases like:

   <table border>
   <table border=>
   <table border="">

...as well as things like:

   setAttribute('border', null);

...and so forth, to work out exactly how UAs generally behave. (We know 
that they don't really implement what HTML4 says, e.g. <input radio> 
doesn't do the same as <input type="radio">.)

If we can get away with it, making <a b> always be exactly equivalent to 
<a b=""> and setAttribute('b', '') would be ideal, as it would simplify a 
lot of things. I've already sort of gone that route with attribute 
definitions, e.g. saying that "the b attribute does bla when it is present 
and bleh when it is not; if it is present it must have the value b", which 
gives one set of criteria for UAs and one set of criteria for authors.

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