[Ktechlab-devel] net-centric ktechlab

[Alan Grimes]

ktechlab was a brilliant, innovative and generally awesome piece of
software for the first decade of the 21st century. Now we're in the
second decade of the century and it's time to raise the bar to the next
level.

Everyone is ga-ga about web-based applications these days. They are
taking perfectly good applications, downgrading their user interface by
five or more years so that they can run in a constraining, slow, and
crappy web-browser.

They're missing the point.

There is no point in running crap in a web browser just because it's the
trendy thing to do.

The point is the *NETWORK*, not the web. Modern software must be network
centric.

There are several major advantages to network centered software.

-> Collaboration.
-> ubiquitous access.
-> real-time sharing of component libraries and platform development.

Basically, these are must-have features for contemporary software.

So what I propose is a client-server architecture with a peer-peer
infrastructure. So if you want to keep some stuff private, you'll need
to run your own server, otherwise it would be optional. (kinda like how
free-civ works.)

The front end will be some kind of game engine. Crystalspace is the
common open source engine, I'm not sure it will support the dynamic
content generation we need. The other interesting engine is OpenCobalt
which is written in Smalltalk. Last I checked it was still definitely
alpha-ware but has shown dramatic progress from earlier versions.

The back end will consist of a parts database and a world-map which will
constantly evolve as people add parts. The server will store the circuit
and the state of the inputs. The client code will run the simulator, but
only on the circuits near the user. (they would otherwise be running
much too fast for any kind of server-side simulation).

I want the client to be 3D, so the schematics can be drawn in 3D. This
is important because nanotechnology will eventually be catching on.

One very practical application would be managing power connections for
IC components and logic gates. Right now we aren't simulating the power
needs of logic gates. These are frequently omitted from schematics for
clarity. For example, a network drawn in the XY plane could have it's
power layout mapped in the X-Z plane...

3D design is also very important for MOS gate design as well as new
layered ICs and PCBs. -- Gerber file exporting will be critical for
making Ktechlab a viable design tool.

Also, it will provide a crossover environment for mechanical
simulations/robotics. (there was some stub code for this in the old tree..)

Now the issue of an "on-line" simulator raises the problems of service
reliability and updating. A peer-peer server infrastructure should
alleviate the cost to the project maintainers for network operation. The
bigger issue is managing software updates to the network. Therefore
there needs to be client-server protocol versioning so that clients can
negotiate with the server. Also, there needs to be a method for
extending the server without taking the whole network down. There are a
number of general approaches for distributing updates. Basically, the
server must be profoundly modular so if someone makes a true
oscilloscope object (as opposed to the chart recorder ktechlab 0.3 had),
that innovation can be integrated and distributed to clients in near
real-time without introducing security problems (!!). So there might
have to be some kind of security check on the server so that, for
example, any code that imports/links against system libraries is
rejected as "prohibited: security hole".

The idea is to create a platform for open hacking that can evolve
collaboratively at a near break-neck speed. If what I propose here is
even half-way successful, it could be totally slashdotted, (survive only
because of the peer-peer server architecture).

There also needs to be some kind of way to manage user identity, so that
access to hub servers can be limited and metered, while users with
private servers can place locks on their own private inventions while
displaying their other stuff in public in such a way that visitors can
copy it but not edit the original.

-- 
E T F
N H E
D E D

Powers are not rights.