Re: [whatwg] Page refresh interface

On Thu, 26 Mar 2015 08:28:37 +0100, Karl Dubost <karl@la-grange.net> wrote:

> Andrea, Simon,
>
>> Le 25 mars 2015 à 23:08, Andrea Rendine <master.skywalker.88@gmail.com>  
>> a écrit :
>>> I think Refresh as an HTTP header is not specified anywhere, so per  
>>> spec
>>> it shouldn't work. However I think browsers all support it, so it  
>>> would be
>>> good to specify it.
>> Indeed. It was Netscape-specific but it's widespread now (that's why we
>> have a <meta http-equiv="refresh"> "surrogate").
>
>
> I was not so sure about the interest of documenting it, but after  
> [digging into it][1].
> There seems to be many Web Compatibility hacks around it.
> [1]: http://www.otsukare.info/2015/03/26/refresh-http-header

Most of what you bring up seems to apply equally to meta refresh, and we  
have a spec for that. That there are issues with interop is why we have  
specs... :-)

The space thing is interesting.
<meta http-equiv="refresh" content="0 url=http://example.com/there" />

The spec says to skip the rest of the string if ; or , is not present  
(step 10,  
https://html.spec.whatwg.org/multipage/semantics.html#attr-meta-http-equiv-refresh  
), but it looks like IE11 and Gecko will instead continue parsing as if ;  
had been seen. That is, remove "Otherwise, jump to the last step." from  
step 10 in the spec to align with IE11/Gecko.

I don't see this in webdevdata, but since webkit/blink have had bugs filed  
about it, I think it seems reasonable to match IE11/Gecko. Filed  
https://www.w3.org/Bugs/Public/show_bug.cgi?id=28338

> Do we have stats on how frequent the `Refresh:` header is on the Web?  
> HTTP Archive maybe?

In webdevdata.org-2015-01-07-112944 I see 58 pages with a Refresh HTTP  
header. I see 4336 pages with meta refresh. 22 have both. Filed  
https://www.w3.org/Bugs/Public/show_bug.cgi?id=28339

httparchive is going to be a bit misleading since the data is fetched  
using IE with the pref to follow meta refresh enabled, so zero-timeout  
meta refresh to a different url will not be in the data set, I think.

-- 
Simon Pieters
Opera Software

Received on Thursday, 26 March 2015 09:03:56 UTC