- From: Ray Whitmer <ray@personallegal.net>
- Date: Wed, 7 Dec 2005 09:23:26 -0700
- To: Johnny Stenback <jst@w3c.jstenback.com>
- Cc: www-dom@w3.org
On Dec 6, 2005, at 2:58 PM, Johnny Stenback wrote:
>
> Indeed. getAttribute() in IE is seemingly mapped to property
> getters on elements, i.e. element.foo === element.getAttribute
> ("foo"). Given that, whether you get null or "" back from
> getAttribute() depends on the element and the attribute you're
> looking for. If the attribute is an attribute that's a string
> attribute, like say, document.body.getAttribute("bgcolor"), you'll
> get "" if the attribute is not defined, but if you do the same
> thing with "onunload" you'll get null back, since that's an event
> handler *property* which defaults to null in IE... Whether this
> ever changed in IE I don't know.
So, we seem know for these historical (Level 0) versions of IE, any
supported string attribute would return a "", not a null.
My claim that IE changed unilaterally after level 0 with respect to
supported string attributes was apparently dead wrong, in part
because it was based on the assumption IE had become basically
noncompliant (apparent claims that the null return was to emulate it).
An implementation which returned null for all unspecified attributes
would seem to only be compatible with IE when looking for js handlers
or other things not stored as strings or other default values or for
unknown/unsupported attributes (debatable in this case if historical
versions did not store unknown/unsupported attributes)?
Is there any attempt to emulate the IE behavior attribute by
attribute in Mozilla?
Also, is there Netscape history that is applicable (from when
Netscape was dominant), that may have played into the interpretation
of DOM Level 0?
Ray
Received on Wednesday, 7 December 2005 16:23:47 UTC