[whatwg] Session history

Hi all,

copy-pasting an IRC conversation about session history.
Seems like HTML5 isn't compatible with any implementation.

[22:58]	<smaug>	Hixie: ping
[22:58]	<Hixie>	hey
[22:59]	<smaug>	Hixie: about HTML5's session history
[22:59]	<Hixie>	yes
[22:59]	<smaug>	I'm trying to understand the "Each browsing context, 
including nested browsing context, has a distinct session history."
[23:00]	<Hixie>	browsing context = frames, iframes, windows, tabs, etc. 
basically i think what mozilla calls a docshell (?)
[23:00]	<smaug>	but in practice if go(-1) is called in a iframe (and it 
hasn't been navigated from the original page), the top level page is 
navigated
[23:00]	<Hixie>	...oh.
[23:00]	<Hixie>	really?
[23:00]	<Hixie>	huh
[23:01]	<smaug>	every browser
[23:01]	<smaug>	don't know about IE
[23:01]	<Hixie>	wow, how did i miss that
[23:01]	<Hixie>	you sure?
[23:01]	<Hixie>	i mean, i'm sure you're sure
[23:01]	<Hixie>	i'm just so surprised
[23:02]	<Hixie>	do you mean that it navigates whichever of the browsing 
contexts was most recently navigated?
[23:02]	<Hixie>	or does it actually move the top one?
[23:02]	<smaug>	Go http://mozilla.pettay.fi/moztests/ then locate 
history_test.html
[23:02]	<smaug>	then press go(-1)
[23:03]	<smaug>	ah
[23:03]	<smaug>	probably whatever was most recently navigated
[23:03]	<smaug>	I need to test some more
[23:03]	<Hixie>	oh that's even more different
[23:03]	<smaug>	but at least the history isn't per browser context
[23:03]	<Hixie>	or at least, .back() and company don't navigate that one
[23:03]	<Hixie>	but some aggregate one
[23:04]	<Hixie>	huh
[23:04]	<Hixie>	that's going to be a huge pain to fix
[23:04]	<smaug>	right
[23:04]	<smaug>	someone was going to start to implement pushState and I 
had to review HTML5's history....
[23:04]	<smaug>	didn't look quite right
[23:05]	<Hixie>	i wonder if i can define a per-top-level-bc history that 
is the union of all the histories of all the bcs that are active
[23:05]	<Hixie>	sorted in the order in which those entries were added
[23:05]	<Hixie>	and then define .go() in terms of that
[23:05]	<Hixie>	and leave everything else alone
[23:05]	<Hixie>	i think that might do it
[23:06]	<Hixie>	i'd have to define the sort order carefully
[23:06]	<Hixie>	and somehow come up with a definition of the "current" 
position
[23:06]	<Hixie>	i'm sure this will be quite exciting
[23:06]	<smaug>	in gecko top level content docshell owns the shistory
[23:06]	<Hixie>	smaug: can i ask you a favour? can you post this irc log 
or something like it to whatwg?
[23:06]	<smaug>	ok, will do
[23:06]	<Hixie>	plus anything else you know about it, like how mozilla 
does it?
[23:07]	<Hixie>	how urgent is it that i fix this?
[23:07]	<smaug>	well, I don't know yet how pushState works with this all
[23:07]	<smaug>	I need to think about that
[23:08]	<Hixie>	k
[23:08]	<Hixie>	i can probably get to this this weekend




-Olli

Received on Thursday, 2 July 2009 13:13:49 UTC