[whatwg] HTML5 - Server-Sent Events Implementation for Mozilla

[Boris Zbarsky]

Ian Hickson wrote:
>> So in practice until page reload, right?
> 
> Or until the URL is added in another <eventsource> element, or 
> addEventSource(), with the same URL, sure.

OK.

>> Why are DNS errors different from under underlying network layer errors 
>> here?
> 
> The intent is for DNS and HTTP 4xx (except auth errors) or 5xx errors to 
> be treated the same way. I'll try to make the spec clearer about this.

I guess the real question is whether it should be treated differently 
from, say, a socket-level error (e.g. "no network interface is up", "no 
route to host", etc, etc).  Seems to me like it shouldn't.

Basically, the current proposed Gecko patch treats DNS failure 
specifically differently from the other possible error status codes 
(socket errors, out of memory errors, user pressing the stop button, 
etc, etc).  If feel like we should treat all such failures (and non-2xx 
HTTP responses, since this is the level above the HTTP impl and hence 
never sees 3xx responses unless redirection failed) identically.

Does that make sense?

>> Also, that still leaves the question of 2xx responses that aren't 
>> explicitly mentioned in the spec.  Those should be treated as 200 in my 
>> opinion.
> 
> There aren't any yet in HTTP.

I know, but I just don't think we want a long list of status code checks 
here that would need to be updated if/when they get added.

> For future 2xx codes, I agree that it would 
> make sense to treat them like 200 OK, so I'll update the spec in due 
> course.

Great.  Thanks!

-Boris