speedtch usbatm.c,NONE,1.1 Makefile,1.7,1.8

[Roman Kagan]

  Hi Duncan,

On Wed, 2005-01-12 at 21:16 +0000, Duncan Sands wrote:
> Update of /home/cvs/speedtch
> In directory phoenix.infradead.org:/tmp/cvs-serv13837
> 
> Modified Files:
> 	Makefile 
> Added Files:
> 	usbatm.c 
> Log Message:
> Playing around with a different way of organising things (like usbnet).

I had a bit of time to take a look at the new file and felt like
expressing my doubts as to whether it was heading in the right
direction.

IMO usbnet is not exactly the best example to follow.  It's a piece of
generic code intermixed with a bunch of device-specific pieces,
unrelated to each other and protected with ugly #ifdefs.  To make it
worse, each device requires multiple #ifdef-protected pieces to be
inserted at different places in the driver.  This, together with the
resulting big file size, makes it difficult and error-prone to add
support for new devices, and to maintain the code.

I believe the current beautiful split-up in usb/atm, with semantically
clean usb_atm library module and well isolated device-specific modules,
is IMO much better than the usbnet all-in-one approach.

What are the particular problems in the current state of affairs that
you are trying to solve with this reorganisation?

Thanks and cheers,
  Roman.