[Gllug] Firefox (was: Tuning the CPU scheduler)

[- Tethys]

On Tue, Oct 27, 2009 at 3:49 PM, Walter Stanish
<walter.stanish at saffrondigital.com> wrote:

> Interesting that you're unhappy with the performance of firefox.

Why do you assume all of my problems with Firefox are performance related?

> Which hardware / kernel config are you running under?  What
> Type of pages are you talking about accessing?  (Anything with
> Java / Flash / embedded PDFs can be much slower than static
> pages...)

I'm running a reasonble spec desktop system (dual opterons, 6GB RAM),
with an out of the box Fedora kernel. But let's address some of the
performance problems for a sec. You're assuming that I'm complaining
about slow rendering. No, mostly I'm complaining about other things:

- Slow startup times. Recent versions are better, but they're still an
order of magnitude worse than they should be (see Opera, for an
example of how to do this properly).

- Slow shutdown times. WTF? All a browser needs to do to shutdown is
save a minimal amount of state to disk and then die (and obviously
wait for the OS to tidy up after it). But Firefox takes *forever* to
shutdown.

- Time taken to switch between tabs. There's no excuse for this. It
should be nearly instantaneous.

- Time taken to open tabs. Go to http://reddit.com/r/pics and middle
click the top 15 or so links. Why is it getting so much slower the
more tabs that are open? Computers are *really* fast, and while there
will always be some slowdown, it's shouldn't be visible to the end
user unless you're memory constrained, and I'm not.

A few non performance complaints:

- When it takes *8 years* to implement a CSS property (inline-block)
that is essential for table-free layout, you know the codebase is
serious rotten. I wasn't expecting a 5 minute fix, but that was
ridiculous.

- Cookie management is abysmal.

- A rogue plugin (yes, flash, I'm largely talking about you here) can
crash the entire browser, rather than just affecting that particular
tab. That just smacks of poor design. I understand this is due to
change in the nearish future, but it should never have been like that
in the first place.

- No command line geometry arguments available.

- Horribly broken when trying to run it on a remote X server.

I'm really just scratching the surface here...

Tet

-- 
“It seems intuitively obvious to me, which means that it might be
wrong.” -- Chris Torek
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug