> 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?
I looked at this briefly; the problem seems to be that the HTTP library
has no idea of what sort of request it's making (<IMG>, typed URL, <LINK>
or whatever). However, I did make the current Accept: pref
"network.http.accept.default" in anticipation of this enhancement.
Because HTTP registers as a pref listener, we _could_ implement this
feature by the code which does know changing the pref before it sends each
request; but I have a horrible feeling this would be too heavyweight and
lead to races. Failing that, this info needs passing down with the request
itself. This is a fair distance (in terms of function calls) and would
need someone who had just rearchitected the HTTP library, for example, to
look into it ;-)
This is now http://bugzilla.mozilla.org/show_bug.cgi?id=82526 .
> Is there a way in principle to say at what length we would stop enumerating
> MIME types, and go do something else (CC/PP?)?
<shrug> I think it's reasonable to say that most people with a machine
modern enough to run Mozilla will have a 56k (or possibly 33.6k) modem.
Embedders may, of course, want to set something smaller - that's, in part,
why it was made a pref. But I think we'd need a very good case to be made
for any more types to be added to the current candidate.
> Just one: you are hereby knighted Guardian of the Accept Header, including its
> reasonable length. Keep it small, comment on and communicate the trade-offs,
> make my gray hairs few! Ok?
OK :-)
Gerv
