On Sep 10, 2009, at 2:30 AM, Robert O'Callahan wrote:
On Thu, Sep 10, 2009 at 8:13 PM, Jonas Sicking <[email protected]>
wrote:
My assumption was always the opposite. For example for <input>
elements we clone the 'value' API attribute, as well as the internal
has-changed-value bit (used for form field restore when going back to
a page).
Looks like Opera and Webkit clone some form control state too (the
text of text inputs, at least). Haven't tested IE, but it seems
likely that interoperability requires at least some cloning of
hidden state, so this does need to be specified somewhere.
I don't particular care what the spec ends up being. For media
elements, cloning some hidden state could be useful, but it is hard
to implement.
Indeed, we seem to copy the following things for form controls: value,
checked state, indeterminate state. We don't seem to do anything
special for any other elements.
The change to clone form state was made based on this bug report: <https://bugs.webkit.org/show_bug.cgi?id=5177
>. The bug report cites Mozilla and IE behavior but does not mention
a real-world site depending on this, though I would presume there was
one.
I think this behavior should be specified in HTML5.
Regards,
Maciej
