- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Thu, 3 Sep 2009 14:34:25 -0500
- To: Bert Bos <bert@w3.org>
- Cc: www-style@w3.org
On Thu, Sep 3, 2009 at 2:14 PM, Bert Bos<bert@w3.org> wrote: > On Monday 31 August 2009, Boris Zbarsky wrote: >> See >> http://limi.net/articles/resource-packages for a very initial >> proposal draft; comments welcome. > > Sort of a side-comment: one other piece of prior art is the DECLARE > attribute of HTML, which allows you to list in the HEAD all the > resources that might be useful to download right away. > > It lists the resources directly, instead of indirectly, as in the Limi > proposal, which means it increases the size of the document more and > thus has a higher risk of slowing down instead of speeding up the > download of the full compound document. > > It exists since 1997, but my impression is that it isn't used. Maybe > waiting for the BODY doesn't actually take that much longer? Yeah, I think that's solving the wrong problem. Or rather, just a part of the problem. Waiting for the <body> obviously doesn't take long; the more important use of it is to prefetch things that aren't used immediately, so you get a responsive UI later. (Of course, I've never seen it used either. Ever. Or even talked about.) The main use-case for sprites is collapsing many resources into one so that they can be downloaded with a single connection. > I also wonder if downloading a zip is really more efficient than > downloading the individual files (assuming HTTP and a browser that > supports pipelining and transfer-coding; it's not the same for FTP and > other protocols). The stream of bytes that comes down to the client is > about the same length as the zip (because the compression is the same). > It contains more HTTP header lines, but the zip contains a directory. I dunno what exactly the effects are of pipelining and transfer-encoding, or how they compare to downloading a single combined file, especially when the numbers of files grow large. > And, if preloading indeed proves efficient, shouldn't it be defined in a > way that applies to all kinds of compound documents, not just HTML? Sure, but those already exist. This is just a particular way to do the single-connection-preloading transparently in HTML. ~TJ
Received on Thursday, 3 September 2009 19:35:20 UTC