KHTML, KJS Unfrozen -- details

[Maciej Stachowiak]

(Catching up on some old email.)

On Jan 11, 2006, at 7:55 AM, Koos Vriezen wrote:

> On Tue, Jan 10, 2006 at 11:50:29PM +0000, Charles Samuels wrote:
>> Koos Vriezen wrote, on Tuesday 2006 January 10 10:27 pm:
>>>>> Seriously, since I've added KJS CPU guard, what will be the
>>>>> replacement?
>>>>
>>>>   I don't know yet.
>>>
>>> Any idea how Safari solves this? (I hope we all agree that a script
>>> never may hang konqueror)
>>
>> I think it'd be nice if Javascripts executed incrementally (like on a
>> QTimer(0) ).
>
> This idea is proposed many times. And it would also help when js  
> has to
> waits for modal dialogs (eg. window.alert()) or async dcop/liveconnect
> calls.

We'd like to rearchitect the interpreter to be able to do this  
(design has been discussed w/ various people). I believe it is  
possible to do in a way that does not hurt performance under normal  
circumstances and it may even be a speedup. Not using the stack will  
make the partially-conservative GC faster and save virtual function  
call overhead, even though it may add overhead in other forms.

> But whatever cool things we imagine, we're at the mercy of WebCore now

We are happy to incorporate good changes. The repository is live, and  
we are highly available by email and irc to discuss design, and we  
process patches pretty quickly most of the time (and are willing to  
grant svn commit access to anyone who consistently does good work).  
We have even given one non-Apple person patch review rights and I  
expect there will be more.

Please think of it as a collaboration rather than being at our mercy.  
We'd like to work together to make the best possible JS engine.

Sincerely,
Maciej