jini vs. ecos vs. lonworks (Score 1)

Here is a comparison from my experience:

LonWorks is interesting, it consists of these
single chip network nodes for ~$5.
The neuron chip is ~$3-4 and there is an
inductor for ~$2 to provide isolation.
The media is a single twisted pair.
Each device is really simple and you need
a central controller to 'administer' the
net. But they tend to get a hefty profit
off that controller portion.

Lonworks is good for well-defined dedicated
tasks. Everyone I know who's tried to do
anything that requires flexibility has
really gotten in to trouble. LonWorks
is best in factory control systems that
do the same thing for 20 to 50 years.
Industrial Automation is a bad space to
play in since factories are so conservative.
Once they are committed to using a vendor,
they tend to stay with it. Lonworks has
serious headaches for the developer.

Using Raw ethernet is not much more expensive
than lonworks, especially when you consider
that cabling costs tend to dwarf both lonworks
and Ethernet components. To make this vision
work requires new wireless protocols.

Jini is an interesting concept. There
are many people at the Media Lab who
have been pushing similar ideas.
They have a whole group dedicated to
Things That Think, and putting intelligence
and networking hidden into everyday devices,
so you can get the benefits without even
knowing that it is there.
This has long-term credibility, but everybody
talks about it while few companies actually
produce. Sun has some incredible concepts
with Jini, simplifying the whole arena of
making devices talk to each other.
This has *long term credibility* since no-one
else is addressing the issue of how to make
protocol design easier for our limited
human brains.
The MIT Media Lab has a special group focusing
on Personal Infromation Architectures
http://www.media.mit.edu/pia
attempting to address these same issues
and solve them. The great thing is, real-world
software can be developed in a research setting.
Real-world hardware requires a company to sell it.
In this space the software is what's hard, and
that makes PIA's work very exciting.

Ecos is just an operating system for embedded
devices. Most people still just write their
own OS from scratch when they need an embedded
OS. The OS starts out as a superloop and
functionality gets added as people need more
features. Finally somebody realizes it needs
to sit on a network and that gets added too.
What you end up with is a *minimal* solution
but at huge development expense. Other
embedded OSes are hard to cut up because
one can't share the cut-ups with others.
Everybody has to re-invent OS partitioning.

Ecos means anybody can strip the OS down to the
bone and share that, leading to a community of
minimal OS implementations tuned for narrow
applications. This isn't revolutionary either
though since a free embedded OS has been available
for some time,known as RTEMS.
At one point RTEMS was selected as the core
OS for http://www.jos.org , trying to build a
free java-based operating system. I'm not sure
if they are still using RTEMS or have developed
a new custom kernel.