[tulip] Re: Clean Interfaces....

Lyle Bickley lbickley@bickleywest.com
Wed, 26 Jul 2000 15:39:53 -0700

Donald Becker wrote:


> This is a bug caused by moving all of the PCMCIA code into the 2.4-pre*
> kernel.
> I can understand the ideal of having every device driver packaged in one
> place with the kernel, ready for instant interface changes with a tap on the
> keyboard.  This ideal might even be feasible if no new hardware came out for
> say, 18 months.
> The current approach is trying to pull all development into the main kernel
> tree.  The alternative is designing clean interfaces that remain stable
> until there is a demonstrated reason to change.  Development for new
> hardware and interfaces should be done in a parallel path, until the updates
> are at least superficially stable and test.

Donald makes an excellent point here.  As a consultant to several
Silicon Valley firms in the area of strategy, I am positive that we can
be assured of increasingly fast-paced changes in the I/O and device
technology.   (And we need them!)

I know there is a rationale for all device drivers to be in a monolithic
kernel.  There is a certain "convenience" to it - and it may "feel" like
we don't know how to do it "perfectly" (we never will).

But if we want to keep Linux current in a rapidly changing environment,
we need to look long and hard at Donald's proposal.  Standards, such as
clean interfaces, CAN increase productivity by parallelizing development

In major projects I have managed over the years, I have found that
defining interfaces definitely takes time and energy.  But in the final
analysis, well defined, clean interfaces prove their value in both
productivity and making target dates.

> Ooops, time to get off the soapbox.  I'm not really a single-issue person.

Neither am I, but this is an important issue.
Lyle Bickley             | Bickley Consulting West Inc.
lbickley@acm.org         | 
lbickley@bickleywest.com | V 650-428-0621
http://bickleywest.com/	 | F 650-428-0599
"Black holes exist where GOD is dividing by zero"