Re: ACTION-961: usefulness of multipart-mixed [recent mobile CSS sprite case study]

Not that this isn't a worthwhile discussion... but am I right in
thinking that we're drifting away from the multipart/mixed discussion
here? And that the consensus wrt multipart/mixed is that whilst it
seeks to address some useful goals (minimising HTTP round trips) and
has a place in certain circumstances, it's not well-supported or
widely used enough to be considered a best practice?

2009/6/8 Eduardo Casais <casays@yahoo.com>:
>
> Stephanie's message brings useful and interesting information to the discussion, but has, in my opinion, not cleared all questions about sprites in mobile environments.
>
> 1) Sprites work well on mobile browsers derived from desktop ones (Safari, Maemo, Opera) and on desktop browsers (Firefox) -- and unsurprisingly these are the ones that are notable in not supporting multipart/mixed.
>
> The following browsers are important in middle to high-end devices, and their status regarding sprites and multipart/mixed should be cleared up: Mobile Internet Explorer, NetFront, Blackberry (versions anterior to 4.6 and those from 4.6 onwards which are based on WebKit).
>
> 2) The examples given in the page "The Mystery Of CSS Sprites: Techniques, Tools And Tutorials" reveal that the practice of sprites in wireline environments tends to be peculiar. Clearly, desktop sites tend to construct sprites that pack together all resource variants that might be used in a site, and then let CSS/Javascript decide on the terminal which elements are relevant and represent them on the display. This is quite evident in the cases of Amazon, Pokemon, TV1.rtp.pt, for instance.
>
> Interestingly, one comment pointed out that Amazon changed its design, with the consequence that nowadays something like half the images in its original sprite (as illustrated in the article) have been eliminated in favour of text elements in the markup.
>
> 3) Given that sprites seem to imply a clever trade-off between the overhead they require and the reduction in the number of HTTP transactions to fetch "decorative" image resources they enable, what we need are hard figures about the compared total latency when using sprites, when not using sprites, and ideally when using multipart/mixed. Taking into account more complex CSS, the fact that one actually ought not to put all required pictures as a single sprite, that sprites only apply to decorative images, etc, how much do we really gain (in absolute and percentage terms) by relying upon sprites?
>
>
> E.Casais
>
>
>
>
>
>



-- 
Future Platforms: hungry and foolish since 2000
work: Tom.Hume@futureplatforms.com play: tomhume.org

Received on Tuesday, 9 June 2009 08:09:56 UTC