At 2004-07-23T16:07:57+1200, Christopher Sawtell wrote:
> I have also come across wierd Modems of anonymous east asian origin.
Another National voter, eh?
> whole business of data on POTS using modems is one huge cludge,
It's "kluge".
> _should_ is the operative word here. So far, my maxim has been "In
> Linus I trust", and it has stood me in very good stead. Thus if I have
> to install an external module I have had a very satisfactory degree of
> success rebuilding the kernel and its modules from scratch so that it
> all links together without problems.
I'll be interested to see how long your blind faith in Linus lasts. At
OLS and on LKML and lwn.net it has been revealed that the 2.6 kernel
will continue to receive feature patches and that any stabilisation work
will largely be left to be handled by the vendors.
CONFIG_MODVERSIONS will take take of kernel/module mismatches.
> On a modern machine, doing that is quicker than drinking a mug of
> coffee. I do my thing while the machine does its job. All over in 20
> mins or so. Look how long people have had to fiddle and curse since
> the 'fest and they are _still_ doing it apparently. Compiling kernel
> sources is nothing to be frightened of.
You must drink coffee slowly. Building a kernel using the
vendor-supplied .config with its plethora of drivers takes a good
half-hour or more on a decent machine.
The end user of "user friendly" distributions should not have to
recompile the kernel just to get a modem to work. They certainly
should've have to review their kernel .config and strip out all of the
drivers that they don't need just so that they can build a kernel before
their coffee is cold.
> May be, but it's really very valuable to be able to recompile driver
> sources so that the linking loader can do its thing without screwing
> up. The modem driver writers are well aware of that fact.
What are you talking about? Commonly a reference to "the linking
loader" would be taken to mean ld{,-linux}.so and friends, which has
nothing to do with kernel modules. With CONFIG_MODVERIONS, the kernel
linker will flat out refuse to load modules that do not match the
kernel. Either it works, or it doesn't, and if it doesn't it's because
the module was built incorrectly in the first place. The linker
Your comments on RPM were hilarious--thanks.
Cheers,
-mjg
--
Matthew Gregan |/
/| [EMAIL PROTECTED]
