Re: document.write from script-inserted external script

On Jul 13, 2009, at 20:40, Travis Leithead wrote:

> Note, we also tested this with IE8+parser fix we've been looking  
> into and our result is:
>
>>> We sometimes succeed and sometimes fail. (I'm guessing it's a  
>>> timing issue with the script download)
>>> When we succeed it looks like Firefox with more whitespace and  
>>> "Kittens! Budgies!" before "BEFORE ... ... AFTER" instead of after.


It seems to me that an HTML5-compliant result shouldn't show  
"BEFORE... ...AFTER" at all. When the load for the external script- 
inserted script completes, the script is run from the event loop.  
Thus, there are no parser stack frames on the call stack and insertion  
point can't be defined. Therefore, the document.write() from 005a.js  
implies a document.open() and blows away the original DOM that had  
"BEFORE... ...AFTER" in it.

The reason why this doesn't happen with the old Gecko HTML parser is  
that it never implies document.open() upon document.write() if there's  
an active parser. Based on Hixie's notes, I *think* the reason why  
this doesn't happen in Opera is that Opera runs the script-inserted  
external script synchronously instead of just scheduling the load and  
letting the execution happen from the event loop.

With my upcoming (though very possibly broken patch), I consistently  
get Kittens!, but I get Budgies!, too, only if I do something unusual  
like bother the timing with a debugger or perform a cached navigation  
back to the page. Also, with my upcoming patch, the page never  
finishes loading, which seems right because the page has been replaced  
with a script-created page that never got document.close() called on it.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Tuesday, 14 July 2009 06:55:13 UTC