|
<![CDATA[
To: Frank Schuurmans <frank@bio.vu.nl>
Hi Frank,
>> At present the Internet is fragmented - we have Email, ftp,
>> News, WWW, LDAP, gopher, telnet etc. Of course, the only
>> one you really need is WWW. You may initially disagree but
>> think about it for a moment - we could have "Webmail" instead
>> of Email. Email could appear as web pages on your server. One
>What about people with only a UUCP connection to the Net.
I am assuming that the Web will become as common as the telephone. I
presume UUCP will die out. What is it anyway?
>> reason this will happen is that punters want full colour
>> with graphics integrated with text and you can't get that
>> with standard Email. You need a webneuron type revolution to
>with attachments you can.
It is very messy, and you need viewers for different file formats. This will
never be acceptable to the punter. The punter needs things in an
exceptionally simple format,
>> do all this but if/when it happens, Email servers will be redundant
>> Going a little further down the line, when the telecoms people
>> become ISP's themselves and each "mini-phone call" is a
>> URL request (you may pay a fixed line rental only), then Webmail
>> will also replace the FAX. The Internet phone should
>Email already is a good alternative to the fax.
Only if you know how to view the graphics attachements.
With weblets there are also many possibilities for analysing the mail as
it is received and perhaps breaking up a long Webmail into relevant parts
and perhaps automatically creating an indexed mini-webletsite from one
Email.
Also, it is better to integrate everything under weblet control.
>> replace the conventional one. Personal computer operating
>> systems will probably be based on HTML pages with suitable
>> extensions to allow programmability, i.e "webneurons". So
>> Webmail will go straight to web pages on your PC without
>> the need of an intermediate ISP. Phone calls will use a
>> PPP system and be in "packets". A spin off is that everyones'
>> PC will become a server, and therefore a publisher. The
>> bandwidth problem will be solved. Speed problems are always
>> conquered, as are storage problems.
>
>Bandwidth is like RAM and HD space you never have enough of it.
Point taken.
>> The others like ftp, News, LDAP etc can also be disposed of
>> by similar methods. In fact gopher and telnet are already historic.
>Telnet historic? You must be joking.
Alright, but do you agree that everything should be integrated rather than
split up?
>> >Yes, I agree with your evolution principle. But wouldn't you say that
things like
>> >Java have happened very fast? Now, I am told to be part of Microsoft
Windows OS!
>>
>> Yes, things like JAVA have happened very fast, but JAVA is realy
only a bolt on
>> which you can take or leave, that runs on top of existing protocols.
>
>JAVA is still in its development phase.
What will it be later?
>The whole Net runs on top of TCP-IP if you don't use it people won't be
able
>to connect to you or they have to add protocols to there computer
There is a conflict here. On the one hand it would be very efficient and fast
to write a weblet in assembler (just on one machine) using say 32 bit codes
to locate each webneuron rather than a URL and TCP-IP. This would be a
stand alone system and would have been perfectly acceptable a few years
ago. On the other hand we now want to grab the extra address space and
distributed computing, provided by URL's and TCP-IP.
How you communicate between machines is a matter which needs very
careful thought in the light of webneurons. The way web browsing
operates is analogous to addressing a memory location in RAM. e.g in
assembler the MOVE instruction just moves data from one place to
another which is what you are doing with web browsing. I don't really
know what I am saying here, except that TCP-IP is probably not ideal for
webneurons, but as you say it would almost impossible to consider
anything else.
>> If you were to try and implement your ideas over the existing http you
will
>> end up with a shadow of your true vision. Your changes are realy at a
very low
>> level, and of a fundermental nature. Either that or you compromise
your plans
>> and end up with "JAVA II" :) which would be much easier to integrate
into the
>> Net, but would defeat the objectives that you have setout to achieve.
>> ---------------------------------------------------------------------[end]
>>
>> If he is right then it may be no good to use Java as you suggest.
>And HTML too off course! You have to 'reinvent the wheel'
>The vision and objectives are still a bit vague to me.
The vision is one standard method of programming at every level using
webneurons which will eventaully grow into a huge network that extends
to and encompasses every piece of software and data in the world. You
will only know of C++ etc by reading about them from some webstore
history library somewhere.
Schoolkids would just be taught the one simple way of programming. This
is possible because webneurons can handle everything the other languages
can, including neural networks, which are currently thought to be the
among the most powerful software tools available (yet are dead simple).
You wrote a preliminary main objective yourself. This is now on the
website as you probably know.
>> ASSEMBLER LEVEL possibilities
>> -----------------------------
>>
>> Having thought about this a bit, it won't be possible to build an
>> assembler based global software
>
>If everbody uses the same CPUs it might be..
Maybe we should develop our own stripped down Internet computer and
attempt to set a standard. I have electronics experience.
We still have the problem of address space. URL's must be built in at
assembler level. This is awkward. Much simpler to use a 32 bit code to
locate pages. This 32 bit code would act as an index to a look up table for
locating the actual 32 bit address in RAM of the start of the webneuron
structure. But we could then use:-
IP address/32 bit code
for global location, assuminmg all computers were 32 bit, which they are
not, so there is another problem to consider. This would be similar to the
current:-
IP address/filename
but much faster.
We need at least four types of hyperlinks on a machine:-
1) Local just the 32 bit code pointing to the caller/target.
a) input
b) output
2) External 32 bit code + IP address
a) input
b) output
In assembler, for a webneuron with many links, this could be handled by
splitting these into four lists within each webneuron. More space would be
reserved for the external links. In assembler the "script" would probably
just be a list of function codes plus data. Each function would process the
inputs and outputs and some would control logic flow. Reading inputs,
and setting or FIRING outputs would be done by functions, and would be
something like "FIRE 1" "SET OUTPUT 4 TO STORE 2", (although not
written like that) where the numbers are indexes to actual links.
I will try and put a picture up on the site of my old plan for assembler level
structures.
We have a further problem because I have heard they are running out of IP
numbers and will be bringing in a larger system where IP addresses take
up more bits.
>> The thought crosses my mind that webneurons could have some things
the
>> brain doesn't. One of them is that since URL's are countless, then
>> webneurons can be countless. So we could make an arbitrarily large
>> software "brain". Whether the other features of a real brain could be
>> understood enough to be duplicated is another question.
>> My guess is yes, over time.
>
>I don't think you can duplicate the brain using a algoritm.
I'm not sure that a weblet "super-brain" would be an algorithm. More like
billions of tiny independent webneuron algorithms linked up by webaxons.
Each webneuron capable of forming countless links with other webneurons
in the network. Thus the map itself takes over from the concept of
algorithm. The "super-brain" would be more of a "linkorithm".
Don't forget that the latest idea for running every webneuron script
continually, gets rid of the FIRE command. The only communication
between webneurons would be the setting of webaxon data, rather than a
more explicit algorithmic link via the FIRE command. This takes us even
further away from conventional computing, much closer to a parallel
system, and may more closely approximate the real brain.
Can you argue that a collection of small algorithms is still an algorithm if
the script for each runs independently?.
>And if it would be possible I wouldn't want my computer to function
>like a human, if it means I get disussions with my computers like
> Me: Gallinago [name of my linux PC], I'd like to finish
> my email to John.
>
> Gallinago: Email ? to John?
>
> Me: Yes, the one I started yesterday
>
> Gallinago: Oepps I've forgotten where I've put it.
>
>or
>
> Me: Gallinago, I'd like to finish my email to John.
>
> Gallinago: I don't feel like finishing email messages
>
> Me: Yes, but we have to.
>
> Gallinago: Lets play chess.
But what about:-
Me: Plexos [the name of of the weblet superbrain], turn part of
yourself into Pamela Anderson.
Plexos: Modify dimensions?
Me: No, just the same.
Plexos: What mood would you like.
Me: Nice and easy.
Plexos: Now just put on the virtual reality headset and relax......
>I'm working on setting up a site. By the way what kind of connection has
>Powernet to the Net, it seems quite slow to me.
From Powenet:-
"we are multi-homed site with 2 x T1, we are
currently using about 384K of the allocation."
They deny they are slow.
I think they are slow. I have tried another ISP which was faster on both
news and web.
Is it very bad? If so I might have to switch ISP.
]]>
|
John's Emails
