[KLUG Members] re: CUPS

[Adam Tauno Williams]

>>> After lots of investigating and a bit of harsh language, I figured 
>>> it out.  Apparently I deleted a 32-bit compatibility library by 
>>> accident, and the printer driver needed it.  Would have been nice 
>>> of it to tell me something was wrong, rather than telling me the 
>>> print job completed!
>> Yep, the current version of CUPS doesn't have a mechanism for filters
>> and backends to return state information to the server; the next version
>> of CUPS has a notification back-channel.
>>> I guess I can take parallel printer support back out of the kernel now.
>> Whats the rationale?  PP support is hardly fringe,  and its presence
>> probably costs you less than a page.
> I suppose I could leave it in, but on this machine a kernel recompile 
> only takes about 5 minutes, and I've never had it fail..  Since 
> everything on a gentoo box (with exception of a few cranky binaries 
> like OpenOffice) is custom compiled, it seems like a waste to run a 
> kernel that's not as optimized as the rest of it.

Ok.  Only I wouldn't call this 'optimization' because you aren't optimizing
anything at all.  The internals of the 2.6 kernel are highly dynamic,  you can
build all the features you want as modules and you are going to get the same
performance then if you peice meal compile features into a 'static' kernel -
because the kernel is not a bunch of static code no matter how much you might
try to make it so.  Recompiling to gain performance only buys you anything in
very specific cases, and then usually only if you know exactly how to 
craft the
compiler and linker options.

And another angle to look at is that if you then want to use the 
parrallel port
for something you are 5 minutes plus a reboot away from being able to 
do so; that is pretty terrible performance in my book.  My OEM provided 
kernel will
get that parrallel device up and running in less than a minute.