[whatwg] Session history and frames in the previous top level browsing context

[Ian Hickson]

On Sat, 3 Oct 2009, Olli Pettay wrote:
> 
> seems like HTML5 doesn't properly specify what should happen if the 
> 'previous top level page' contained frames and go(-<some value>) is used 
> to navigate back.
> 
> Browsers also do different things. A pretty simple testcase is here 
> http://mozilla.pettay.fi/moztests/history/Start.html Just click the 
> links until 'B iframe 2' is loaded. go(-2) should apparently load 
> previous top level page, and in its iframe it should load the previously 
> loaded document, not the document, which iframe element in the top level 
> page refers to. Section 6.10 doesn't seem to specify that behavior 
> properly.
> 
> Perhaps even more interesting case is go(-3). Which one is the right 
> thing to do, IE8 behavior or FF3.6/Opera10? And if IE8's, why its 'Back' 
> button behaves differently.

IE8 actually behaves exactly like what the spec says, it seems (and it 
does seem to match it's back button, unless I'm missing something). As far 
as I can tell, the spec isn't ambiguous here.


> And this all is still just about static pages. Things get more 
> complicated when (i)frames are added and removed dynamically.

The spec actually doesn't explicitly cover anything to do with how to deal 
with navigating subframes when a Document has been evicted from the cache 
and is reloaded during session history traversal. I'm not really sure 
exactly what to do about that. It's not clear to me what the right 
solution should be. Does anyone have any proposals?

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