[exprla-devel] Re: [XPL] - WorldOS - framework for distributed computing

[reid_spencer <ras...@re...>]

--- In xpl-dev@y..., Jonathan Burns <saski@w...> wrote:
Richard Anthony Hein wrote:

>
>
>      Everyone,There is something called WorldOS that is a
>      framework for distributed computing that could be useful to
>      XPL.  If you read some of my other posts you will know why I
>      like the idea of using such a framework. It includes "a
>      simple TCP server, an XML oriented application server, and
>      tools for peer routing similar to Freenet or Gnutella.  The
>      application server uses a new protocol distinguished by
>      pluggable transports, XML message content, and asynchronous
>      messaging."It's opensource, so that's good, and it's in the
>      early stages of development, but does function.  So we could
>      get our input in there once we have an architecture for XPL,
>      that would make WorldOS even better for our purposes.
>      Anyways, take a look at
>      http://www.worldos.org/worldos/index.php3 to get the
>      details.Richard A. Hein
>

So happens, I've just stumbled onto WorldOS myself.

This was just after I'd gained a clue what Gnutella is.

(Gnutella is a search engine for files. Load a query for some file
by name. The query fans out among all the Gnutella users your system
can find, and all that THEY can find, recursively. With luck, download
access to the file(s) you're after will eventually be passed back to
you,
and show up in a window. Meanwhile, your system is picking up its end
of other users' searches.)

And Napster, which does the same kind of thing for MP3s music files.

And FreeNet, which gets you onto a web of, shall we say process or 
user
identities, allowing exchange of data on an IP-independent basis, more
or less untraceable by the usual TCP/IP methods.

Furthermore, this is after I have been reading and hearing for some 
time

about:

PunkNet, an architecture like FreeNet.

Linda, a network OS and language in which "tuples" circulate through
data stores on widely distributed machines, being gathered and 
processed

by some kind of content search.

Seti@Home.

The Morris Internet Worm, and others of its kind.

To cap it off...

Check out Kurt Cagle's  slideshow  on VBXML. It's couched in general
principles,
but if I have it right, Kurt's vision is of free-floating,
call-'em-by-name
XML documents, which encapsulate LOCATION and PROTOCOL, so as to 
present
the
casual user with a web of data bound to data, leading wherever you 
want
to go
via search-engine interfacing. No more searching for a site, going
there,
accessing the FTP or whatever, and finally accessing the data you want
by name.


This is looking like a paradigm.


Now I completely agree, Richard, that we should investigate WorldOS -
and maybe
make common cause with it. See, WorldOS will evidently conduct any XML
documents.
So if we maintain a layer within XPL development which is Pure XML - 
no
processing
primitives, no COM hooks, no binding to external mechanisms - then
WorldOS will
function as a transparent distributed mechanism for that layer.
Technically, it's
a potent tool for us.

But sheesh.

Is anybody as unnerved as I am, about the implications of "data gas"
webs like this,
especially when people like us are going round, endowing them with 
full
processing
capability?

Call me conservative. I spend a lot of time fuming about how free
markets have
diffused the consequences of actions beyond the point where anyone can
be imputed
responsibility for them. Collectively, we've committed ourselves to a
world society,
while disabling the corrective checks and balances that a society 
needs,
against
the concentration of power.

The Web is the same, in fast-forward. It's this fast, not because of
communication
speed, but because there's so little power involved in publishing and
scanning
information, that it's relatively harmless. But already, the Web is
facing a massive
contradiction: intellectual propery funds it, but its consequence is 
the
devaluation
of intellectual property.

What happens when information is completely dissociated from 
authorship?

When programs are completely dissociated from authorship?

When worms can be released in perfect anonymity?

When you can't tell the good worms, like Seti@Home, from the bad 
worms,
or from
worms like Napster which sidestep around intellectual propery?

When there is no such thing as a trusted source?


Well, that's the New Economy Political Agenda, anyway.

On a more technical plane, what bemuses me is, how do we find things 
in
a data gas?

Answer: We refer to them by content rather than location - and we use
personal
or site caches to nail down copies or shortcut addresses. At a guess, 
if
or when
we go and implement an XPL data gas, we will be using XML tech to
implement
the Universal Resource Access mechanism we need. XPath on steroids, 
man.

<spooky>

I've got this vision, of an ocean of little XML thingies - parse trees
on the loose.
They leaf out in typed nodes. Spotlights sweep across the ocean - 
people
and their
sundry bots, paying attention, and thereby supplying CPU power to the
thingies.

Thingy awakens. It finds a phone. Good morning switchboard, I'm 
looking
for stuff
with these types and content and timestamp and authentication (that's 
my
leaves).
Oh, you got some - good.

Instantly, thingy is linked up with others of its kind, making a
temporary Big Thingy.
And so on recursively. At some point, all the leaves are filled in, 
down
to primitive
elements, and there's some XPL present, which activates its XSLTs, and
the whole Big
Thingy goes to work - on borrowed CPU power, from a multitude of
sources.

Presently, Big Thingy has bred a bunch of children, and put them on 
the
phone to
the people and bots who originally requested whatever services. They 
get
linked
up with GUI thingies, which deliver the service to the end user. End
user satisfied,
spotlights blink off, thingies go back to sleep. But some middlesize
thingies are
bigger now, or more current. They contain a little bit of information 
on
the
late transaction, cached away, ready for the next time.


The whole thing has more than a passing resemblance to brain function.

Furthermore, it has a close structural resemblance to John Koza's
Genetic Programming
paradigm.

Here again we have parse trees on the loose - Lisp expressions. Here
again we
have the free-floating population. From the population, a quantity of
elements
are selected, and activated. They are programs - feed them to a Lisp
interpreter
(spotlight) and they run. This can be done in parallel. It can be done
like Seti@Home.
It's a data gas.

The outputs of the programs are collected, and scored. Depending on 
its
score, each
activated element is given breeding instructions - drop out, or
replicate yourself,
or split apart, or split and recombine with some other element, on a
type-compatible
basis.

A new spotlight comes on. In its warm glow, the elements follow their
instructions.
The population is modified. Collectively, it now contains more 
elements
which are
fitter, more likely to score high. And it contains many elements which
were parts
of fit individuals, and statistically likely to have contributed to
their success.

It's an evolving data gas.

</spooky>


Kltpzyxm!
Jonathan

His wit never failed him, and he laughed himself to death
over a book of the dying words of famous men.

- Thomas M. Disch, Camp Concentration.
--- End forwarded message ---