Re: XMLHttpRequest test suite suggestion

Boris Zbarsky wrote:
> If I follow correctly, relative URIs passed to open() should be resolved 
> relative to caller (whatever that means; more on that in a separate 
> mail). That's not actually precisely what Gecko does right now; probably 
> a bug in Gecko.
> 
> In particular, if you have documents A and B at two different URIs that 
> can talk to each other:
> 
> 1)  Document A creates an XMLHttpRequest object
> 2)  Document A sets an onreadystatechange listener on it.
> 3)  Document B calls open()
> 
> the base URI will be document A as far as I can tell from the code.  
> Worth adding a test to the test suite for this and similar things where 
> two different windows both work with the same XMLHttpRequest object.
> 
> Note that each step here needs to be done off a timer or something to 
> avoid the ambiguity as to what "caller" means that I alluded to earlier.

See also http://www.w3.org/mid/441E69F1.2030709@sicking.cc

/ Jonas