To: Frank Schuurmans
Hi Frank,
Really sorry not to have mailed you sooner. I got
a new computer last week and of course it didn't
work properly. Took a lot of time up. Also I
put a lot of hours into the website. Restructured the
Email and added a "Main Topics" section. This was because
Spire E-zine wanted an "information pack".
I have also put up all the backlog of Email.
>I understand webaxons to store the data that get
>transfered from webneuron to webneuron. If each
>webneuron uses a axon for storing for example an
>image (let's say 1Mb, which isn't big for a true
>color image). Now we have a small weblet which
>consists of 20 webneurons. each time a webneuron
>fires the data gets copied to a new axon (This
>costs time). By the time we get to the last
>webneuron me have 20*1Mb in memory.
This is something important that I have thought
about quite a bit myself, i.e. how do we handle
the data? I think you're right - it can't make
sense to pass large data around like that. Too
much of a drain on resources. Also, the brain
does not operate by passing large data around
either.
>My idea of data is that it should't be linked to
>webneurons staticilly but dynamically: a webaxon
>just travels from webneuron to webneuron and each
>neuron modifies it.
But the idea of a webaxon is that only the caller
can change the data in it? If it travelled then
I don't think you can call it a webaxon anymore,
plus the Assembler model would have to be scrapped.
If different webneurons want to modify data in a paricular
webaxon they can be linked to the webaxon caller
webneuron, and can then make changes via the script.
The data doesn't need to move at all. In other words
the travelling is done by forming links rather than
by moving data. So the assembler model can do
everything you want. It doesn't need changing.
It is correct and can handle all tasks.
I am not against your ideas, but I am concerned
to have a standard model that we can work from
as soon as possible so that we can then address
the difficult decision of which platform to use.
>The overall model would thus become:
>
> a map is send (by the user or a webneuron) to
> data (a webaxon) the webaxon 'travels' from
> webneuron to webneuron using the map.
Would this be like chemicals travelling to
the DNA or from cell to cell?
>The map could contain webneurons on other computers,
>in which case the data will be send to the other
>computer which modifies it and then send it back
>(or not depending on the map). This sending of
>the data could be done using any protocol there is
>(http ftp or whatever) as long as both computers now
>that the data which is send has to be processed.
>(a very simple solution would be to set up a cgi-script
>on both sides. A way to view those scripts is a
>receptor model (If this is not clear to you know
>i will explain it later on maybe a got place would
>be on the web so I can use pictures more easily).
I am keen to set up CGI scripts at both ends like you say
and try a few experiments. I think I will get Linux
on the new computer.
>You might be intrested to know that real neuron also
>use receptors: one neuron produces a chemical
>(called a neurotransmitter) and excretes this at the
>synaps the other neuron has a receptor (in fact many
>receptors for different chemicals) for this chemical.
Yes, I know a bit about neurotransmitters. Isn't this
quite easy to model in software though. Doesn't it basically
operate like a trigger if the level of neurotransmitter
gets high enough. This could be modelled by an easily
programmable simple threshold function.
>This model of receptors could be extended to include
>not only the communication between computers but also
>within computers: we could have a web for handling
>files (a device driver for harddiscs) which has also
>uses this receptor model. This would mean that you
>could easily replace a part of the system with another
>part just make include a receptor for the data travelling
>in your computer. (the way i like to view this is like
>a organ transplantation)
>
>What do you think of this idea?
I find it interesting and connected to some thoughts I had
about "grafting" weblets. If you wanted to join two weblets
you would have to specify how all the links connected.
It is a problem I never solved.
How is your website doing and what will be the URL?
I would also be interested to know what you are going to put
on it.
|