Testing browsers on low-speed connection

I tested how popular browsers behave on a thin stream of telephone-modem connection. Behave, I will say right away, mostly impolite, refusing to perform their main duties - to load the user with graphics.

The last place is Firefox3 - it didn’t upload as many as 19 images (from ~ 50). Then Safari with the result -11. Closes the top three IE6 with minus six.

"Only" two images missed Opera . And K-Meleon won, without losing a byte on the road.
')
Strange results ... Well, okay, you can not load the resource (everyone has different timeouts), but why not finish the work about which it is known that it is not finished yet?! (Returning to the original page, everyone suddenly remembers that something is underloaded and urgently runs to load.) To achieve the result, there are all possibilities, and this should not depend on the line speed, the number of streams, or the use of pipelining . However, it depends.

Either the developers of all browsers overtake, or me. Let's find out who and what :-))

(Anticipating) comments in the spirit of “now everyone has megabits” are not welcome.

Technical details ...

I will not indicate the loading times, because it has completely lost its meaning among such results. From the interesting, I would only note that Opera worked for a minute longer (5:17) the winner-Chameleon.

The opera managed to push (max. 4 threads) to the “excellent” state. With the fox, the same focus in about: config failed, although it was once possible. How to tweak wayward IE and Safari I can not imagine. (In any case, we are talking here about the default settings and even that these settings should be unimportant.)

It is interesting that K-Meleon is based on the same Gecko engine as Firefox (two months older), and the results are so different.

Testing conditions:

There are approximately 50 “child” download objects on the test web page .

Versions:
Opera 9.50
IE 6
Firefox 3 (Gecko May 29)
K-Meleon 1.1.5 (Gecko, April 6)
Safari 3.1
I have no others

Everything under Windows XP SP2

Speed: 33.6 something there

The cache was cleaned every time. Browsers installed again, cleaning the registry and profiles.

UPD: These are the results of the first run. There was a second with the same winners and losers. “Everything” alternated to prevent the BG from letting one browser fall into the bad bar. Yes, and not so it does not matter how many runs there and how stable the line is, understand! Browsers do not know how to live in a situation of limited resources - this is what a ficus is.

UPS: I myself use a fox%)