- From: Reinier Kaper <rp.kaper@gmail.com>
- Date: Thu, 12 Sep 2013 11:45:13 -0400
- To: Anselm Hannemann <info@anselm-hannemann.com>
- Cc: Simon Pieters <simonp@opera.com>, "public-html@w3.org" <public-html@w3.org>
- Message-ID: <CAAz96Oui9+PU+T+GSrwVg3VHWszbgPehSHos+CRQOVjTx_vmBw@mail.gmail.com>
On 12 September 2013 11:37, Anselm Hannemann <info@anselm-hannemann.com>wrote: > On 12.09.2013, at 17:35, Reinier Kaper <rp.kaper@gmail.com> wrote: > > >> >> Why can't we do something like <img src="some-lowres.jpg, >>> somewhat-higher-res.jpg, very-high-res.jpg" > and then use CSS selectors >>> to >>> decide which src to apply? >>> Very rough example: >>> @media (min-width: 18em) { >>> img { >>> src: 1 // This would be an index based number and if not found, >>> default >>> to 0 >>> } >>> } >>> >> >> Because browsers want to start downloading the correct image before >> stylesheets have been downloaded and parsed. > > > That's why I proposed that browser download only the first source file by > default. This will create extra overhead for when the CSS has been parsed > (and an alternative source needs to be downloaded) of course, which might > be an issue, but at least there's no tie-in between mark-up and CSS and > there will always be an image to display. > > > > Just on the last: This is what we need to avoid. Double-download is a > no-go for a responsive images solution and in such a case wouldn't solve > the attempt to reduce image load and improve performance. > And there's no way for a browser to detect *if* CSS will be parsed? So you'd end up with two scenarios: 1. No CSS will be parsed on this page, therefore load src[0]; 2. CSS *will* be parsed on this page, therefore don't start downloading image resources yet; 2.1 When CSS parsing takes place: detect which media query is used and load the corresponding src accordingly. The issue here would be when the browser scales up/down, but then alternative resources should be loaded asynchronously and only fully loaded sources should be replacing the previous one. Maybe I just don't have the technical knowledge, but to me it seems this is something that the browser would ideally deal with.
Received on Thursday, 12 September 2013 15:45:41 UTC