On Tue, 03 Jul 2001 17:11:30 -0700 esteemed Asa Dotzler did hold forth
thusly:
>
> Debug|chofmann's browser buster. It doesn't report directly but if
> enough people run it with a talkback build it will help to solidify the
> mean time between failure numbers. I've been running it regularly for
> quite some time. Back in November of last year I was thrilled when I
> hit 200 pages and now I'm bummed out if I don't make 1000 (on linux).
> We're definitely getting better.
Is there any kind of browser buster that would navigate different browsers
from different vendors in order to generate comparative data on stability?
I wonder if IE is scriptable to be controlled to go to a series of sites.
Running Moz and IE side by side with a scripted series of sites would yield
some useful data. But since IE's performance couldn't be reported back by
talkback we'd need another way to report back the results.
So I'm wondering how to go about doing this. I'd like to take a stab at
making a utility that would do this to run IE and Moz side by side and then
to make a report (initially that would have no way to be automatically sent
back) to describe how far each of them got.
A few questions:
1) How does the chofmann's existing browser buster choose the series of links
that it visits?
2) Is there some list of the top n hundred sites visited? Or top n pages in
those sites that are stable URLs that get visited?
3) Would it make sense to have a page scanner that would go to big sites and
then link to some links off of them and then send the two browsers to these
links in succession?
4) I think the main testing program should try to load each page itself
before telling each of the browsers to do so. Once it succeeded then it would
direct each browser to load a page.
5) Can a separate process tell Moz to go to a page? (preferably without
telling Moz to open a new window but it'd be nice to test with multiple
windows opening as well)
6) Can Moz use IPC to report back that it finished loading a given page?
I think a tool to compare browsers would need to run as a separate process
that did not crash when a browser crashed. So hence the need for some form of
IPC via perhaps a named pipe or a socket connection to the browser.
