[wplug] Tell me something good

[Ingimarson, Darin]

Larry,

> Late last year there were a considerable number of articles in the trade
> journals about how Linux and Java were going to be the dynamic duo of
> embedded designs.  Indications were that even if Linux never makes it to
the
> desktop, the server and embedded market would insure its future.  Even
with
> of all this positive press, one of the leaders in the embedded Linux
> markeplace, Lineo, has just about closed its doors this week.  They have
> cut their staff from 320 to 72 and have had their Houston office
padlocked.
> Lineo provides (provided?) the operating system used in the Sharp Zaurus
> PDA.

I think I can actually contribute something from a position of 
experience here... I think what you are seeing is the continuing
shakeout of the embedded industry in general. Everybody is
building less, freezing hiring, and not taking on any new projects.
Companies that depended on the "hockey stick" graphs that sales
wonks were flashing around a year and a half ago are now trimming.

IMO anyone who has used Linux in an embedded system in the past 
for valid reasons (i.e. a need to get at the kernel internals,
rapid deployment on COTS hardware, etc.) will probably continue
to do so. 

However, in my opinion these types of developers are more apt to
"roll their own" embedded kits (which is what I did) once they 
have a good hard look at what companies like Monta Vista, Lineo,
and Lynux Works REALLY offer for the $500-$25K they want for
developer seats.

What you really get for your $$$ is the following:

1) Some sort of crappy ROM image / disk image builder (these things
   typically just round up the binaries and place them onto a 
   mounted file full o' zeroes and formatted with the ext2 fielsystem...
   you can then burn this to a ROM, Flash, or write it to a floppy
   to boot your system, not hard in principle).
2) Often the company will insert a proprietary boot loader into the
   system to screw with anyone trying to use a different kernel -- 
   i.e. one that they have not tested in-house (this is their way of
   cutting support costs). 
3) Engineers for tech support.

Anyone who has the wherewithall to play with hardware at the 
embedded level certainly has the ability (and probably the desire)
to have a build environment that is somewhat configurable -- possibly
more configurable than what you pay for. That,in my opinion, is the
strength of open source.

I can attest to the inflexibility of so-called "shrink-wrapped"
embedded and real-time OS vendors whne you go to them with even 
a slightly differing system (most embedded/RTOS vendors have an
"approved hardware" list that they want you to use with the 
software). I would know, RTOS kernel development was my first job
after university.

> Companies are realizing that free isn't free.  Porting Linux to their
> embedded processor is technically difficult, taking too long and costing
too
> much.

The simple fact is that it is just as complicated to develop an
embedded application with an open-source OS as if you were to 
use, say, VxWorks or similar. It just depends on whether or not
you need/want to pay for OS support (essentially that is what
you are buying when you buy one of these commercial packages).
At that point, you just choose the OS for which you want to pay
for having a squadron of app engineers on the other end of a
phone to help you. 

Those who think that "Open Source" and Java are magic bullets to
rapid deployment and reduced time-to-market with embedded systems
are kidding themselves. Development of a deeply embedded system is
complex and needs to involve developers with intrinsic knowledge of
hardware and software interactions. Simply sitting a newbie down
with a compiler (C, C++, Java, whatever) and a piece of circuitry 
and saying "Make this work... you have 6 weeks" is an assurance
of disaster.

I can't agree with your comment about porting to embedded processors
though. The real work in porting an OS lies in supporting the 
peripherals on the board. In a truly custom platform, you will still
have to support MMUs, I/O chipsets, bridges, etc. If you develop
your own board, or use an oddball, you have to write this stuff
either way.

> Redhat has also had severe cuts in its embedded staff, and this morning I
> receive a notice that after only 10 issues, SSC will no longer be
publishing
> "Embedded Linux Journal".  They site "it is no longer viable as a
> stand-alone publication".

Trade magazines like "Embedded Linux Journal" are funded primarily
by advertisers, and if you look at the articles, you can see that 
a really healthy portion of them are in fact WRITTEN by those
selfsame advertisers (which brings up a tangential point -- be 
careful what you read in articles like those, bear in mind who
is grinding the axe). With decrease in sales on behalf of the 
advertisers, no wonder the parent publication is pulling the plug.

Anyway, the point I am trying to make (minus the coffee induced
babbling) is that anyone who had a reason to use embedded Linux
probably will continue to do so. And yes, anyone who jumped on 
the bandwagon will be shoulder-rolling off selfsame bandwagon
about now.

Linux will continue, irregardless of the death of a few companies
along the way. It's the nature of open source. Someone else will
pick up the ball along the way. Guaranteed.

	-darin

P.S. Go to http://www.quantapoint.com/ppi_solutions.html  The
laser scanner pictured there is an embedded Linux application 
on a PC/104+ stack.