[extropy-chat] Firefox 2 [was: The End of Science]
robert.bradbury at gmail.com
Sat Oct 28 14:32:51 UTC 2006
On 10/27/06, Russell Wallace <russell.wallace at gmail.com> wrote:
> I have some ideas. Number one on the list is to stop programming in mega
> macro assembler (which is what C++ is)
That much you got right. They should have used a *real* systems programming
language, namely C, as was done with the Linux Kernel and the X Windows
system. But it doesn't matter if you are programming in C, Perl, Python or
Java -- one should elegantly handle resource limits -- particularly running
out of memory.
(except obviously for small embedded systems where mega macro assembler is
> what's called for). Which in turn means we need to stop fretting about a few
> gigabytes of memory consumption :)
Ok, I'll stop worrying. I'll go get an AMD Opteron with a 64-bit address
space and put 16GB of main memory on it and add an entire 160GB hard drive
for swap space (probably only set me back $1000 or $1500). Then I'll put
together a list of about 200,000 URLs, stick them in a session file and tell
Firefox to open it. (PubMed has millions of abstracts so this shouldn't be
a big problem). Since I'm guessing I'm running about 1MB per URL after
about a week of waiting for the pages to download and be rendered Firefox
*should* crap out the same way it craps out now when I set the virtual
memory ulimit to 150MB. This is because it will run out of swap space and
the unhandled out-of-memory allocation errors will blow it out of the water.
Software either works reliably or it doesn't.
Software either works efficiently or it doesn't.
It has little to do with complexity or the programming language involved.
It mostly has to do that most of the focus is on bells and whistles and
trying to compete with I.E. rather than making it reliable and efficient.
I strongly suspect that this is because the core Firefox developers include
more than a few former Netscape employees that want to get even with
Microsoft for taking away the Netscape browser crown years ago.
They have failed to learn what Google clearly demonstrated. Make it work
well enough and make it work fast and you can beat those who happened to be
first and had a greater market share (e.g. Altavista). 
1. I'll note as an aside that this is what made Oracle the leader in
databases back in the 80s and 90s as well. A strong focus was placed on the
TPS performance results. Fortunately there was a well defined SQL
"standard" and less emphasis was placed on bells and whistles to
differentiate one from the competition.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat