- From: Amos Jeffries <squid3@treenet.co.nz>
- Date: Thu, 08 Nov 2012 12:05:12 +1300
- To: <ietf-http-wg@w3.org>
On 08.11.2012 08:26, Tiffany B. Brown wrote: > On 11/7/12 10:17 AM, Eitan Adler wrote: >> On 7 November 2012 13:04, Tiffany B. Brown <tiffanyb@opera.com> >> wrote: >>> Hello all, >>> >>> I've recently submitted an RFC for a new HTTP header: >>> Device-Stock-UA. >>> Feedback is welcome. >>> >>> http://www.ietf.org/staging/draft-brown-device-stock-ua-00.txt > > Brief aside: I've been informed that I should have called this an > Internet Draft (because it is). For the sake of following the > discussion, I am keeping the original subject line. > > >> It isn't at all clear to me that this solves the problem: >> >> - the stock services may have been changed. just because a browser >> comes with a device by default doesn't mean it wasn't removed. >> - the capabilities of the third party browser may differ from the >> capabilities of the stock browser > > This header should be used in conjunction with feature detection if > used at all. Perhaps that's worth mentioning in an implementation > notes or similar? Feature detection of *what* exactly? that is important. critical even. Being informed about a piece of software which is not related to the transaction underway is useless. And I say this with my experienced web developer hat on. For example; when opening a Z format object in Browser 'X' do I need to know the image processing capabilities of browser 'Y', or the fact that Picassa versus Photoshop are installed? No. I need to know that the browser X has a Z interpreter and is using it to handle the processing of the response I about to send back. If browser X requests an image object, I still don't need to know that browser Y was installed, or even that the imaging software is installed. Just that the browser has *an* image processing component *it is making use of*. There is nothing more annoying to both developer and user than to have a browser send explicit mention of capability Z, only to display that Z format response using a text processor and dump raw binary out at the viewing user. > >> - user agent shouldn't be use for capability detection in the first >> place > > Agreed. However, this is precisely how the user agent string is being > used. This is particularly true for sites and applications that > support hand-held devices (mobile phones, tablets, etc). Developers > are relying on the user agent string to determine which image and CSS > files to serve. > >> - what does the (stock) user agent string have to do with "other >> software that may be running in addition to the user agent"? >> > > Developers are also using the user agent string to infer things such > as the presence of audio and video codecs, plugin availability, and > whether the OS permits file system access. *This* is the problem. Your header is making zero progress on fixing it. Those websites are ignoring the Accept* headers content where such codec and media type support is listed in annoyingly great detail. And instead depending on the browser agent name string. Amos
Received on Wednesday, 7 November 2012 23:07:17 UTC