Argh, having trouble replying, this new Mozilla 0.99 is really buggy :-(
Well, Hurd's theoretical thread model is hard to beat. It can spawn
threads on remote machines / cpu's with no problem. The complexity
is what slows us down. Cray et. al. has been running a Hurdish-like
OS for a long time, but us poor Hurd hackers don't have the resources to
work 24/7 on the project.
Multithreading is good. It makes no sense to run a single-threaded app
on Hurd if a multithreaded model is better in any way at all. We currently
prefer poor performance and "good design" rather than non-portable cheap
assembler hacks. After all, computers are getting faster and more numerous
all the time, but CPU's have changed radically!
L8r!
Sam Dennis wrote:
Some time around 7 o'clock PM on March 12, a Tuesday, Sammy Mannaert wrote:
"B. Douglas Hilton" wrote:
It cheerfully complied with no problems, loads up the default
window in X, and then kind of goes out to lunch if you try
to open a file or url.
you can also compile dillo by changing defines in src/dns.c
(outcommenting #define G_DNS_THREADED)
this way dillo will not use multithreading :)
Unfortunately, asynchronous DNS is not the only thing dillo uses threads
for.
IO/file.c contains other code, and no alternative method.