[whatwg] Declarative unload data

[Charles Pritchard]

On 5/7/2012 12:39 PM, Glenn Maynard wrote:
> On Mon, May 7, 2012 at 2:05 PM, Jonas Sicking<jonas at sicking.cc>  wrote:
>
>> As far as resource usage goes, it doesn't seem like a bid deal to send
>> off a request at some point when the network seems quiet.
>>
> I could see interop problems if people use this to send something when the
> user navigates out, and then the page they navigate to tries to use the
> results.  It'll only work if the request happened more quickly than the new
> page.  That's probably going to happen most of the time (if they're going
> to the same server), so I could see people inadvertently depending on it.
>
> This probably can't be prevented entirely (the only way to do that would be
> to wait for the final request to complete before navigating, which is no
> good), though browsers delaying the request if the network is busy would
> exacerbate it.

How about limiting it to shared workers?

Shared workers are already in the background: we could fire off the 
end-game event right before shutting the shared worker down.
Authors ought to be looking at a shared worker if they're sharing state 
across pages.