On 12 Jun 2008, at 14:33, Anne van Kesteren wrote:
status and statusText currently MUST throw INVALID_STATE_ERR when
there isn't any status code or text respectively sent by the
server. HTTP/0.9 includes neither: Saf, Fx, and IE all return 200
and "OK", and Op returns 0 and "". There isn't actually any issue
with the state, so throwing an INVALID_STATE_ERR makes little
sense. Also, a fair number of servers manage to omit the
statusText, and that should just return "OK" (per Saf, Fx, and IE).
I'd say that it should only throw if the state is UNSENT or OPENED.
I think it would be better if HTTP defined what clients should
assume (200 and OK most likely) in case the response data does not
include it. Your HTTP parsing specification could do this for
instance.
I think that we should have this in XHR. Basic summary is that Firefox
and Safari default to 200/OK; Opera defaults to 0/"" (but does not
throw INVALID_STATE_ERR); IE is inconsistent and sometimes gives 200/
OK or -1/some-random-value-from-the-HTTP-response. I think we should
probably just spec in XHR that 200/OK should be returned when there is
no status-code/reason-phrase.
We're fairly close to interoperability (as IE already sometimes does
it), and nothing matches the spec currently at all, I think it should
be put in XHR and not wait until my HTTP parsing spec, and waiting to
see if anyone will actually implement that.
--
Geoffrey Sneddon
<http://gsnedders.com/>
