(update - the author of the fix wrote)

" I thought I would just chime in here, since I wrote the fix ;-)
So far I have heard of two users having this Connect button problem. I honestly have no idea how that is happening, and it is much harder for me to debug than the panics (more on that below).

The fix solves a race condition in the serial driver. The race is actually present on single processor machines as well, though it is much less likely to occur (this is true of many SMP bugs).

I do not actually have a multiprocessor machine, and do not have easy access to a dialin connection, so I wrote the patch entirely on my Pismo, without actually connecting to the net. I want to thank Hiram and everyone else who gave me such great feedback.

There is one more fix I am aware of that may cut down the frequency of system losing its PPP connections. I am not absolutely sure it will work, but I will take a shot at a little later when I have some time.

Without an actual machine my odds of debugging anything that does not panic are not very good, since it is much more difficult for me to get any useful debugging data (I don't have source code to Internet Connection, so to diagnose why the button doesn't work would require me tracing it at runtime, on an affected system). What can I say, if I can't reproduce it, or read a backtrace of it the problem is a lot harder to solve.

A friend offered to let me use his DP box to actually work on this, that should help a lot.

The fix *should* not cause any new problems, but it seems like in limited circumstances it is. I am working to solve it.
Louis"

" Hi Mike,
I wrote to you in the past about a particularly nasty bug in OS X which seemed to effect the vast majority of dual G4 owners using PPP. The bug specifically effects dual G4 machines running OS X and connecting to the net via PPP and the internal modem. Every time I tried to connect to net using OS X, the session eventually ended with kernel panic.

I had tried contacting Apple directly numerous times about the problem. I contacted them by phone, by email, by posting feedback to their feedback address and by posting messages on Apple's discussion forums. They never once acknowledged that a bug existed let alone gave me any hint that they were working on a fix. Rather than considering that there might be a bug in OS X, they had me strip my machine to its factory original state, reformat my hard drive and reinstall OS X. When this didn't solve the problem, they sent me a replacement modem to try. When this failed to correct the problem-- they were basically speechless.

About a week ago I saw a message from Louis Gerbarg, a Darwin developer, and he offered to take a look at the kernel backtrace data that results from these crashes. I sent him my backtrace data along with the backtrace data of many other dual G4 owners.

I am very happy to report it appears Louis Gerbarg has not only found but also fixed the bug. The bug was in Apple's serial driver (AppleSCCSerial.kext). This driver is part of the Darwin Opensourse code. Louis tracked down and fixed the problem with driver and then compiled a new version which he has made available for download. This is the version of the driver I'm using now and I have yet to experience a kernel panic.

See this message board thread titled, "Multiprocessor modem users needed (test fix)"

Topic: Multiprocessor modem users needed (test fix)
http://forums.macnn.com/cgi-bin/Forum3/HTML/004852.html

Louis' new driver is testament to his excellent programming skills and also speaks very well of the Opensource concept. If I, and many others, had to wait for Apple to fix this problem, who knows how long it would have been before a fix was publicly available. Instead, thanks to his generosity and skill I am now using OS X to connect to net without a kernel panic in sight.
Thank you,
Hiram"

Note: if you're still having problems with this update, try disabling the PPP Echo Packets option in OS X (previous tip here).

Update - although general feedback seems to be very positive - a reader from france noted a problem that apparently is not typical:

" Don't use the fix for PPP/internal modem.
Since I did it, I can't use the connect to internet tool !

I'm a french user and now, even after reinstalling the original driver, the button CONNECT has no effect at all !
I suppose I have to reinstall my system !
Special thanks to Hiram
Patrick Jamet"

I can only guess this case is rare, (perhaps not installed properly) otherwise there would be more problem reports.