[Pc_Support] AnandTech on Intel DuoCore vs USB

[Bryan J. Smith]

Damien McKenna <dmckenna at thelimucompany.com> wrote:
> Intel DuoCore vs USB, Anand is the referee:
> http://www.anandtech.com/mobile/showdoc.aspx?i=2693 

The problem _always_ has been and _always_ will be that USB _lacks_
standards.  It was designed by Intel (IC) and Microsoft (software
host) to put _all_ of the brains in the end-device.  That means the
"host controller" has virtually _no_control_ over how the device
behaves.

Without those standards and without that control, everything is
pretty arbitrary.  Left to the device driver provided by the end
device.  That results in conflicts, incompatibility, total lack of
the host being able to pre-empt the device, etc...

Now there have been _some_ post-USB release standards created.  But
most  are not followed, or only a small subset are.  So this doesn't
surprise me one damn bit.

By the same token that Linux does _not_ work with a lot of USB
devices, for those USB devices Linux _does_ work with, it's
_well_implemented_.  This is typically because A) the hardware must
conform to many post-USB release standards/APIs and/or B) the drivers
are written by the Linux USB team/developers, which _include_
advantaged management features.

This Linux v. Windows 2000/XP on USB is a _repeat_ of the Linux v.
Windows 9x on PCMCIA/CardBus issue.  For Windows, Microsoft just
relied on vendors for PCMCIA/CardBus controller and end device
drivers -- and they often _conflicted_ with each other (before NT
standardized many things).  The Linux world built a _proper_ set of
standard host interfaces and drivers were written meeting those host
interfaces for both the controller and end devices.

So, again, it doesn't surprise me one damn bit that USB devices
basically _control_ Windows, and not vice-versa.  Ironically enough,
Apple (who has a BSD licensed kernel and can sign NDAs -- a major
plus over the Linux community) is far better of an example than
Microsoft on how to "tame" USB.  Until Microsoft starts taking
responsibility for building a _proper_ USB subsystem, instead of just
leaving drivers to the end-devices, this will continue.

Just my $0.02 ...


-- 
Bryan J. Smith     Professional, Technical Annoyance
b.j.smith at ieee.org      http://thebs413.blogspot.com
----------------------------------------------------
*** Speed doesn't kill, difference in speed does ***