Brendan Eich wrote:
>
> Gervase Markham wrote:
>
> > > I'm already concerned about performance effects of the above candidate for
> > > modem users -- these go out on every image request, right?
> >
> > Every request. The above is 167 bytes - which is 0.05 seconds on a 33.6k
> > modem.
>
> You mean 0.05 for 28.8? 33.6kbps modems are symmetric and a bit faster.
> Anyway, we're on the slippery slope. 10 images, half a second, etc. What
> happened to the idea of tailoring Accept based on the requesting tag?
Which seems like a very good idea. After all, Moz really doesn't accept
text/html (even if sent) when requested from an IMG tag?
Perhaps there could be another pref specifying what the image accepts
are. What about JavaScript <SCRIPT SRC=>? You probably don't want to
accept images or xhtml+xml in there.
If it is possible to enumerate possible sources of request (e.g.
"general load", IMG tag, SCRIPT tag), then one could write a set of
prefs specifying what each source of request accepts (e.g.
prefs...accepts.img [i don't know what the naming of such a preference
is] would be "image/png, image/mng, image/jpeg, image/gif;q=0.1"). Then
the default Accept header would be a sum of all such
source-of-request-specific Accept header parts.
Wouldn't such a mechansim significantly increase the usefulness of the
Accept header?
- Oleg
