Looking at problems in a
new way.

Project Oxygen at MIT is looking at human
centered computing, this is a good place to start
but I would ask the question should we not think
about throwing away all existing paradims?  I
am currently working through (slowly) Stephen
Wolfram's "A New Kind of Science" one of the most
interesting things so far is that he has approoch
the problem of understanding complex systems from
a completely fresh point of view. I cannot tell if
he is right or wrong, I will leave that to time
and people smarter than I. the major contribution
I see so far is that he has challenged
preconveived notions of how things should be.

So is Project Oxygen looking at things new or
just re-packaging what we have, while I admire
their goals, I cannot help but think that they are
starting with a broken platform and they actually
need to go deeper into the fabric of how we design
computer systems.

Take Seti@Home as an example, they have been able to use millions
of hours of shared computer time, but
unfortunately very few problems are ameanable to
this type of analysis. There are several grid
computing and peer to peer systems today that try
and attack the shared computing issue but they are
struggling, because, IMHO they are constrained by
the fundamental architecture we have built.

I think this is where Stephen Wolfram has got
(at least the approach right), instead of beating
your head against a wall trying to brute force a
solution (I used to work in fluid dynamics so I
understand a little of the issues surrounding the
complexity of modelling), take several steps back
and try and find another approach.

So what would be a new apporach to the
computing problem, and where will it come
from?

The other side of the
problem....

If we change the information model, moving from
a centralized database model to a more distributed
model how do we handle persistant and recovery.
Very few systems are set up to have the same
degree of backup and recovery as relational
databases, the flip side is the need to centralize
systems due to management complexity.

How to change the information model and yet
still provide a high degree to reliability and
accessibility?

Why are we
limited....

To only what we can access directly, the web
opens up access to all sorts of information, but
it is still a fetch model rather than a push
model. Subscribing to RSS channels is better but
the channel list and information is still manually
selected. There have been several attempts to
create agents that live on the desktop to learn
your interests and then go get information for you
but they have been limited for several reasons,

• limited to one machine - I want my agent to
  move with me (Radio has this problem - I want my
  Blog editor to be mobile)
  
• Unable to differentiate between various
  aspects of life, personal, hobbies, and work
  
• Need to interact with me and get guidance
  

So how do we turn our information models into
push rather than pull?

Information does not want to be free it just
needs to be understood!

Information
models.....

If the problem with computers is that the
information model is wrong, by this a mean is that
it is too complex to access and use information. A
good example may be the current problems the FBI
and CIA have had relating information, a lot
existed but all the relative information was never
put together or even a fraction of the relevant
information was assembled.

Is one of the problems we are trying to ensure
a perfect solution rather than get it close enough
to get the job done. Designing information models
to be totally predicatable i.e. put them in a
static store, typically a relational database
takes too long and by the time it is done the
information is out of date. There was an article
on Slashdot, refering to a new search engine
Kartoo.
While several other efforts have tried this
approach and I am not sure this one will be
successful it does point the direction to a new
way of finding and navigating information. What
kinds of user interfaces could be buiilt on this
model?

How to get
there..

Like any major change in technology it will
probably require a fundamental shift and a
dramatic new idea that will be simple and widely
available. There was an interesting article in the
Economist this week regarding innovation in a
capitalistic system. One of the key points is that
in a free market economy, the benifits of
innovation are widely dispersed, so it is not only
the originator of the new technology that benifits
but society as a whole.

It is intesting to consider what pressures
cause that spark and if we are there yet, 
for the economic benefit to reach deep into
society, of the information age, it must be easily
accessible to everyone. Today hardware and
software is uniformly hard to use, the browser has
helped a lot as it has actually forced designers
to create simple software that does not require a
great deal of knowledge to use - this does not
solve the problem as most of use using Radio (with
great respect to the people at UserLand)
configuring it is not user friendly.

The problem goes much deeper than the user
interface, it does to the fundamental information
model our software is based on. In turn this may
be influenced by the current computer
architecture.

What is the nature of the information model
that makes software so hard?

If I could just access the resources when I
needed them...

I would just be able to access the computing
resources I needed, without having the need to
manage and maintain a computer. How would the
economics of this work, is it even possible.
Similar to network storage models the economics
are not possible as the equivalent of Moore's Law
for storage costs makes storage cheaper faster
than the capital costs can be recovered in usage
fees.

What would this model look
like....

If I did not have a
computer....

but had access to computing resources the same
way as I have access to the phone network and the
power grid, would life be better? Not sure there
are many arguments that can be made that the power
grid is too centralized, I think the power grid
will start to evolve when cheap fuel cells (or at
least some form of cheap and efficient energy
storage). Computers on the other hand are too
distributed and hence the costs of management are
too expensive and complicated. This may sound like
a rant for the return of thin computing but it is
not - if I start a rant I will clearly mark it
with <rant></rant> tags :-).

Very much like the ASP market some good ideas
were masked by many bad implementations. Taking
off the shelf enterprise software and turning it
into an ASP offering IMHO was very dumb, it was
not designed to do that and the only cost savings
that could be acheived was install and
maintenance, additional costs were added by the
need to have secure connections to a remote
facility, which also added latency and bandwidth
costs. ASP's like SalesForce on the other hand
developed applications that were tailored to the
market and are becoming quite successful.

So back to computers - thin client computing
was not successful for several reasons, I think
mainly it was an idea not only before its time but
one that lacked the appropriate software
infrastructure to support it.

Are Computers too hard to use...

Almost a redundant question, we would all
probably say yes but for many different
reasons.  It has been really eye opening to
me seeing my various family members struggle to
connect to the internet and exchange information.
I have started questioning all the activities I
was doing to maintain and run my own
computers.

Should we start not as engineers but rather as
niave users - for most of us this is very hard, I
wrote my first program in Algol in 1975, since
then I have programmed almost in every language
and operating system, ( I missed Cobol and
mainframes....). Living through the problems of
smart but technologically unsophisticated users
brings home the problems of getting connected. Are
we trying to do too much, very like the VCR - too
many functions, with a user interface that
requires continued reference to the manual. As
engineers we are driven by adding more bells and
whistles rather than driving to simplicity. Why
was the Palm so successful - not the breadth of
features but for its simplicity. Unfortunately the
handheld market is heading in the same direction,
people are staring to compete in terms of features
not in terms of simplity.

So should we be thinking of a new design
approach to defining and delivering
solutions....