Tech Fusion Outline: Organising the World's Knowledge.

With the introduction of the Internet, vast amounts of information became available - and, rather than help people of the planet to become useful in a globalised world, it has deluged them. Peeking through the morass of software and hardware is the occasional light (hopefully not an oncoming train). This article will outline those technologies. briefly, for later expansion.

The "Executive Summary" is that for computer technology to be useful, we need modular portable hardware with wireless mesh networking as well as standard internet access, and for the software applications to sit on top of distributed and peer-to-peer technology.

None of the technology outlined here is new (in fact, some of it has existed for many decades): it's just not being brought together. It should be pretty clear that in the current world climate, there is some degree of urgency to making this "Tech Fusion" happen.

As I aim to describe these technologies in more detail later, I mainly want to outline them and give hints as to their relevance, as a first priority. The goal is simple: *actually* provide people with a means to articulate their thoughts, needs and desires, and to be able to communicate those thoughts needs and desires to whoever can fulfil them. world-wide.

When this happens, we will have brought about - literally - a "World Age of Enlightenment". literally, because the literal definition of an "Enlightened Society" is one in which everyone in that society is "useful".

Databases: improvements for distributed - and modern - uses

The design and principles behind Databases were thrown together nearly forty years ago, and have not been updated since. In the article titled "The Vietnam of Computer Science", the next forty years have been spent endeavouring to make a two-dimensional concept (rows, columns) fit into the much more useful and generic "tree" or "free-form" structure concept of Objects and their inter-relationships (Object Relational Mappers - ORMs).

it doesn't fit.

One company spent twelve years wading an ORM back and forth from a hybrid of c++, SQL and Stored Procedures, changing the percentage of code in each programming category over the years depending on the whim of the ever-changing management. At one point, they had nearly 100% of the code - generating Enterprise-grade forms and reporting for data management in the Oracle Stored Procedures, and twice in the company's code development they had nearly 100% of the code in c++.

A great deal of time is wasted on writing code in as many languages as exist on the planet, when in fact the addition of OO features to a database would make that almost entirely unnecessary.

Additionally, significant amounts of time are spent developing "database replication" technology, which is the old "client-server" model gone badly wrong. What is actually needed is for applications to take into account a distributed and peer-to-peer architecture, and for the databases to actually help them do that.

Also, Databases were designed when procedural programming languages were the norm: we've since moved on a bit, but Databases haven't. Whilst most modern programming languages are Object-Orientated, so-called "4th Generation" Database technology is either unheard of or prohibitively expensive.

Here is what is required to update Free Software databases to be "useful" in a distributed context, and they all focus on being able to store "object relationships":

• Addition of Primary Indexing on 16 bytes of (un)signed chars known to most as uuids, but that's just a uchar[16]. Those people with a knowledge of cryptographic hashing functions and peer-to-peer DHT will immediately recognise why this would be useful in a global environment.

• Writeable views are absolutely essential. I've outlined before why, and many of the algorithms I came up with have been reduplicated in sqlalchemy.org. It is absolutely essential, as it becomes possible for people to think and access the database in 2nd Normalised form yet the underlying structure is entirely in a completeley different 2nd, 3rd or even 4th Normalised form. The executive summary: we need a "Meta-SQL" compiler or interpreter which takes in any type of SQL syntax and any queries, and "maps" them to a different form, then makes the queries in any *other* type of SQL syntax.

• Inclusion of vtable and inheritance concepts in SQL databases. This isn't new: it's a well-understood essential and integral part of every single modern compiler and interpreter. Each "object" has a set of "function pointers" and each object which inherits another extends its "function pointers". In SQL terms, that means that each TABLE requires to be identified by a UUID (which can be stored in a special control table) and an "INHERITS FROM tablename" concept added (oh look, another special control table). Then, Queries are not just a matter of looking in one table for the data, it will be necessary to continuously refer to the "inheritance control" table to merge in additional results from the inheritance stack, down to the "base" object. i mean table. The issue of there being differing "objects" returned with varying order and size of row results can once again be solved by providing the row names and orderings in yet another special control table. Yes, this may look complex, but if you think it's complex, I'm sorry to have to point this out but you *really* don't know how absolutely awful it's been for SQL programmers over the past 40 years: do read that article "The Vietnam of Computer Science".

Peer-to-peer Distributed Technology

The "client-server" model is fine for centralised control: it's not fine in an environment where there aren't any servers. Or, where the technology is expected to operate stand-alone as well as integrated as part of a larger group.

• Distributed Hash Tables are becoming increasingly common: the underlying principle is to create a 128-dimensional hypercube, with the 128-bit (16 byte) "hash" as the "address" in the hypercube. Nodes with only one-bit of difference between "addresses" (hashes) are considered to be immediate neighbours in the hypercube. Search algorithms make repeated queries to increasingly-distant nodes, asking for keys which have progressively less and less "bits" different from the hash actually being sought. This technology - DHTs - must be a standard component of every single modern programming language and O.S.

• uPnP, STUN and STUNT: firewall-busting and NAT traversal. the thorny issue of running out of IPv4 space isn't unique to IPv4: pure IPv6 Gateways are still going to present exactly the same problem. So a series of services - a networking layer - is required that provides these "Traversal" technologies "as standard".

• An ISO Layer 3 sockaddr scheme is required that is a standard for peer-to-peer services. It would probably suffice to make it 16-bytes (plus a 16-bit port, maybe?). struct sockaddr_p2p or similar. Behind this simple sockaddr would be the technologies listed above: DHTs, STUN etc. Implementations of services which provide the technology already exist - for example, cspace, gnunetd and i2p - they don't however have the kind of level of "acceptance" that is required to make it easy to port any application - easily - into their framework. Also, after reviewing many p2p applications and frameworks, it would help to have a SOCK_DGSFW - "Datagram Service which Stores and Forwards". ioctls specifying the priority and the time-to-live will at least be required.

• A peer-to-peer "Naming Service" which will replace the need for DNS. DNS itself should be included in the PPNS as one of the sub-services offered. It would in fact be pretty easy to put zones and sub-zones etc. into a PPNS. The concept of PPNS is borrowed from DCE/RPC's "name service", which, funnily enough, provides exactly what is needed (see FreeDCE and also the release from The Open Group of DCE 1.2.2 under the LGPL). It even has a central (global) naming service and the information is stored in an X500 directory - search for isode [which would have to go: replaced by a peer-to-peer directory which would need a distributed database ohh look are we getting the picture here? :) ]

  Also, it's important to recognise that good "Naming Services" have, as an absolutely essential feature, the means to "group" providers of a service together. This simple feature means that, under a particular service, many "providers" can register - and be found. DNS itself has this very concept, where you can register many nameservices or mailservers - but typically you have a central "control" over how the DNS zone is managed.

  There are many examples of "grouping" Name Services - NetBIOS (rfc1001.txt and rfc1002.txt) is an example of one of the most widely deployed (and most widely misunderstood) proper peer-to-peer naming services that contained this strategically important feature of "group registration". nmbd, in samba, implements NBNS. Unfortunately, NBNS implementations are restricted to "WAN" deployment, and do not scale to millions of systems (without a little bit of tweaking, and I understand what's needed).

  A level of security in the registration of who can provide a particular group service is required, which again brings us on to requiring a distributed peer-to-peer secure authentication service.

• A Peer-to-peer secure authentication service is required which is based around the concept of "who your friends are". Whom do you trust, and to do what? (e.g. "I trust this person to talk to me, but their computer contains viruses and they keep sending me spam, so I dont "fully" trust them.") Instant Messaging has had this concept for decades: it keeps all the people selling sex away from your children.

  The IM "buddies" concept is a bit like Advogato's very own Trust Metrics. A more formal version of Trust Metrics, which has PKI Digital Signatures on top of it, is "keynote". Keynote implements the (missing) part of the equation that Raph Levien outlines in his paper on Trust Metrics, and advogato itself implements the missing part of the equation that keynote achieves [actual real-world use :) ]

• RPC technology needs to be much better used, developed, and understood. DCOP is a good example of how *not* to write an RPC mechanism (i heard it took 20 minutes). DCE/RPC was written by legendary visionaries whom i have the utmost respect for - yet DCE/RPC was written as "Enterprise" middleware, not as "World-wide" middleware. There is a plethora of RPC technology out there, yet people still make the mistake of thinking that it is okay to ignore communications errors in their RPC applications. Correction: Object-Orientated RPC technology is needed. We're in the 21st Century, now.

• Mesh Networking needs to become "the norm". There are several implementations: Zeroconf, IEEE 802.11.4 (Zigbee!), TETRA, gnunetd's VPN service; in i2p, mesh networking is even built-in, to make it attack-resistant. Note: Zigbee "as is" isn't so useful due to the limited range of the standard, but the peer-to-peer networking which is described in great detail in the standard *is* useful.

Skype is the perfect model which every "useful" application should strive for (except that it's not free software). Skype is simple, useful, easy to install, easy to use, does the job, doesn't go wrong even when being attacked by the Telecoms and ISP Industries, and provides seamless voice, messaging and video communication when all other VoIP options are effectively living in the stone age. Why? because skype solved the problems by using peer-to-peer technology.

GoogleTalk *almost* does the same job - except that, if you've examined the source code, you will notice that it is designed to integrate into google's infrastructure, not for the provision of completely independent peer-to-peer communications. Notes indicate that if you cannot contact the network, you must go via a "proxy". No such "proxy" service is provided in the libjinglep2p code: the design is incomplete.

Virtually every free software peer-to-peer application that you encounter contains its own RPC mechanism of some sort, its own DHT algorithm, its own naming service, its own search capability, its own networking capability. All of this technology needs to become ubiquotous.

Hardware: Modular Design, Mesh Networking, Hybrid Fusion of Purpose

The hardware I describe here exists or has existed - in commercial and volume production today; in research labs; in spy networks operating since the late 1950s - and yet, and yet, it's not yet provided in a single device. The reason: competition. There's no money in it, or it's too powerful, or cannot be controlled. Well, with the planet falling apart around our ears, it's time to get this hardware brought up-to-speed - *fast* - before it's too late.

• Small, compact, portable, interchangeable. That's the design requirements. A hybrid machine which is sufficiently powerful to be a computer, a communicator (instant messaging, SMS, voice and video), a multimedia platform (video and audio) - just absolutely everything.

• Touchscreens. Essential. Mice aren't really portable.

• Laser keyboards are "in". Keyboards are the number one World Health Organisation health hazard.

• Interchangeable screen/case. With a compact snap-together design, the modules should be easy to fit into whatever "package" that the user desires. For example, the same module could fit into a form-factor case which is a standard mobile phone: 1.5in screen plus buttons. The same module could fit into a smartphone case, with a 4in VGA or a 3in QVGA touchscreen. The same module could fit into a Tablet PC case, with a 1024x768 or 1200x800 touchscreen, USB extensions etc. and a big battery. The list goes on.

• Micro-projectors from companies like Light Blue Optics. 1024x768 or 1200x800 projections is perfectly feasible, readable indoors at a projection distance of 15in. That is good enough to be a replacement for a desktop TFT screen. Also, a recent invention is a 140 lumens lightbulb that uses plasma, can fit into a space approximately 7mm cubed, and only uses 1 watt. That's as powerful as a streetlamp. 6,000 degrees centigrade inside. only one watt. This little lightbulb makes micro-projectors actually viable: the Mitsubishi PK20 mini-projector, which uses the world's only *commercially* available projector integrated circuit (from Texas Instruments) is only 25 lumens.

• The 2009 Intel 45nm CPU - somewhere around 1.5ghz, somewhere around 0.1 watts, somewhere around $USD 6 because a single wafer will have over 2,000 processors on it. Integrated Video. Integrated NorthBridge. Almost perfect. Makes you wonder why Intel sold their PXA processor design, because that would be a far better CPU for the target market of UMPCs and Smartphones, and, as the entire mobile phone industry revolves around the ARM CPU, switching to x86 is utterly painful. Perhaps that's a good thing. Alternatives are hard to find (or a compromise): Samsung's SC2440 series; the AMD Geode LX-800. Intel, as the monopoly, really is the only one offering technology powerful enough. Mr Intel, you have an enormous responsibility!

• Modular Communications The hardware needs to have tiny snap-on communications modules, so that a traveller going from one place to another can simply obtain what they need and immediately be able to access any wireless networks (point-to-point or mesh) in their area. GSM/GPRS/EDGE. WCDMA. HDSPA. WIMAX. TETRA. ZigBee. 802.11abgn. Even Low-Earth Orbit Satellite (for the future, that one...) Ultra-Wide-Band is the one that *really* needs attention, however, as the hardware is simple, low-power, resistant to disruption, and can easily transmit even on a GHz carrier over distances of several kilometres. Ultimately, however, a module is needed which can be reprogrammed (see gnu radio project). A powerful SIMD engine on an ARM core would be perfect, to directly process and create the raw R.F. signal.

• 8-way phased ceramic antennas provide direct, accurate and near-impossible-to-disrupt communications. The array of ceramic antennae do beam-steering, which must be done at the R.F. level, not at baseband. Short of a nuclear strike, it's damn difficult to disrupt. And, because the beam is directed, much less power is needed.

• Aluminium Batteries. Europositron.com have a design of an aluminium-based, sealed 1.5 volt rechargeable cell that is FIVE times more powerful than a NiMh cell. Unlike former aluminium batteries, the aluminium compound is on the anode not the cathode, and so the battery does not turn to slush when discharged: instead, the liquid compound crystallises. Nanotech materials are needed in the manufacturing.

• Voice-based authentication - voice "fingerprinting". 140 metrics can be taken from sentences which are sufficient to provide 100% accurate, 99.9% reliable identification - even with a single bar of GSM signal. A two-way authentication mechanism is also part of the invention, which is often what "secure" bank systems forget about. There's no point providing authentication mechanisms that can be "Phished"!

• Voice recognition - a chip has been invented which, when combined with some Content-Addressable Memory, can instantly recognise phonemes and look them up into words. in any language. Currently, the technology can only do about 1,500 words - but that is more than enough to run most people's computing devices.

• Environmental Sensor Array should be included as sta ndard, with sensors detecting temperature; humidity; Carbon Dioxide levels; Carbon Monoxide levels; artificially manufactured sapphires that can be used to detect poisons in water supplies; even a miniature spectrum analyser (anyone know if this is feasible?). The purpose of this miniature equipment: collecting environmental information to protect people from hazards and, more importantly, the planet we live on. For example: a world-wide uncensored incontrovertible distributed database of the levels of Carbon Dioxide in populated areas would finally open people's eyes.

Knowledge, Ontologies, Parser Auto-generation (Reverse-Engineering)

Organising the world's information is a trivial task of applying simple, well-understood distributed algorithms onto increasingly-large amounts of hardware. "well-understood" does not necessarily mean "well-liked". For example, I heard somewhere that google told people that it uses brute-force search instead of database indexing (which didn't go down too well on slashdot).

Organising the world's knowledge involves contextual inference of meaning. Pattern-matching. Classification (known to the Web 2.0 community as "tagging") - but automated. Classification of classifications involves levels of recursion that quickly damage most people's brains.

It's *not* trivial. There are a handful of people in the world who understand how to organise knowledge, and they mostly work for Intelligence Agencies, and they are absolutely committed to their work. There do exist commercial applications: their price tag usually starts at five figures, and comes attached with contractors whose daily rate is in the four figures range, because you simply won't "grok" their tools.

The key to understanding "Knowledge" is in the Vedic scriptures, which are *not* religious texts: they are an expansion of quantum mechanics functions. Many readers will have difficulty accepting this simple statement. The only thing that I can say to you is: it's taken me my entire life so far to understand enough to be able to state what I have, and it's only in the recent few weeks that I've begun to fully comprehend the significance of Vedic knowledge, and its relevance, and the parallels with computing technology. You therefore have two choices: trust my words, or work it out for yourself. In the meantime, the planet goes down the toilet, and we don't have a replacement planet.

The goal of "Organising the World's Knowledge" is simple: to make it possible to search for any topic, and to immediately be connected to the world's leading authoritative individuals in that area, and their work.

To achieve that, tools are needed which can pattern-match similar-looking information, whether that information is in text, images, voice, video: anything. Some formats are going to need more computation than others, but it is not an insurmountable issue. Many of the formats will need accompanying text written by humans to provide context - in fact, *all* of the formats will need to have their context taken into consideration.

Ironically, all of the components needed already exist - even as Free Software: they just haven't been integrated, because their significance hasn't been recognised. Here are the components which, when integrated, will provide the framework to organise the world's knowledge. I'd like them to be implemented in a peer-to-peer framework.

• Vedic Science and Modern Science are beginning to merge. Quantum Mechanics. Dempster Shafer Theory. Theories of Consciousness. "The Secret". "Down The Rabbit Hole (Quantum Edition)". Ironically, in order to truly prove the parallels between Vedic and Modern science, organising the world's knowledge is, undeniably, a way to connect the dots.

• Ontology Classification (aka "Tagging" in Web 2.0 terminology) is a key requirement. The AMOS project (part of the E.U. funded 5th framework) was completed in 2005. The AMOS project implements a pattern-matcher which can match similar pieces of text, and it was used to match up source code of free software projects (like... all of them). Whilst the AMOS code doesn't actually tell you *what* it's found is the same as that which was found in another text file, it can at least *find* similarities. You then have a front-end tool which presents the "finds" to a human, who then gives "names" to the matches, along with descriptions. However, ultimately, it's possible to perform matches on the classifications, and the descriptions of those classifications, along with the context of the many "finds". This is where it gets horribly recursive and awful and I don't want to go into details.

• The GPLv3 "annotator" which was developed to show comments made during the license's revision stage is a good example of a knowledge-based front-end.

• Auto-parser-generation - yes it's possible. The authors of python-hachoir have written a reverse-engineering-assistance tool which I believe implements Dempster-Shafer's algorithm. Amazingly, it's 250 lines of code. The code as-is doesn't create recursive parsers: actually, you don't want it to. But, what you do is you then perform recursive pattern matching on the subsections that are sub-divided at the level above.

• Parser development assistance and Reverse-engineering tools . GoldParser. OllyDbg. IDAPro. SoftICE. All of these tools assist in producing formal definitions of, and understandings of, data.

• Context-based tools LEO. FreeMind. LEO is the most promising, but FreeMind is mentioned for completeness.

• Genetic Algorithms have always been useful: learning from nature's example. Even over ten years ago I remember an article which described how random variations in algorithms, run thousands of times and tested for correctness, could result in speed improvement techniques being "rediscovered", such as loop unbundling and loop invariants. It doesn't have to stop there: you just have to trust that it works (for a given level of paranoid testing "trust"). For any requirements, you just have to define the tests correctly, and "trust" that the process will end up with the "best-fit" results, even if you don't understand what happened.

• Parsers which accept tree-structures as input , where a special class of input is a tree of left-associative trees with one "character" in each node. To retrieve data from trees is pretty easy: you create a class which has a "GetNextToken" function. Ordinarily, in a "normal" parser, the next character in the stream would be provided (solution for LALR parsers left as exercise for reader). However if the "GetNextToken" was a recursive-depth tree walker on a Tree-node.... So you could even walk an XML tree as input - or even the results of an XPath query, or even the output from a computer program could be "streamed" to a Parser (wow!). This is the principle behind LEO: LEO provides a framework in which this kind of incredibly powerful parser concept actually works.

  Note to people who are familiar with LALR parsers such as flex and bison: I'm acutely aware that the technique of "GetNextToken" and "lookaheadtoken" is already supported - it's just that it's not recognised as being "useful". As part of a knowledge system, I assure you it most definitely is.

  Note also: an LR parser which supports BNF form can be written in FORTH in under fifteen lines of code (including comments). It's so fast that it is completely unnecessary to expand it to have LALR capabilities.

Laws and Licenses

This is perhaps the most thorny of all the issues. Corporations have bought so many laws, where possible, that in some ways it is easier to forget about the countries where this has been done. However, the point of "World Knowledge" is that the entire... world is involved. So - here is a list of things that need to be resolved:

• Dissolve or beef up the United Nations. The jury's out on this one. Read "First and Last Men" by Olaf Stapledon. Stapledon points out, in the very first few pages of Chapter 1, the ineffectiveness of the League of Nations and its replacement by the "United Nations". Well - the same thing is happening to the United Nations: it's becoming irrelevant. Why? because of the "veto" voting system, and also because of the fear of upsetting those Nations primarily funding the U.N. - no really important decisions can ever actually be made. Belonging to the U.N. should be like jury service. Read Orson Scott Card's "Ender" Series - look up how Graff has to be ever so careful. (btw - note to american readers: ignore Baxter's introduction in which he agrees with everything that Stapledon says apart from how Stapledon must have been wrong in his assessment of the U.S.)

• Reestablish Sovereignty. As Sovereign Nations look to the UN to act, it's quite clear that each Nation really does need to get its individual act together, for matters that are their National Responsibility. That will including standing up to being bullied by other Nations, for example by not handing over individuals who carry out reverse-engineering or security breaches (note to people who are bristling at this: you shouldn't allow access to your systems across National boundaries, should you! why are your systems accessible across borders? what are you up to in another Sovereign State such that some of your citizens require access to your "secured" servers, across National borders??)

• Dissolving WIPO - The World Intellectual Property Organisation needs to be shut down - or its job transformed into one which *protects* sovereign nations from the interference of intellectual "property" hoarding. I consider the concept of "Intellectual Property" to be "slavery". even the name says so! Intellectual. Property. Intelligence. Owned. Information. Enslaved. It's got to stop. We're not a bunch of savages. any more.

• Dissolution of Patents. See above. Also: see the documentary "The Corporation". Even the reasons why the Patent system was created is flawed: hoarding of information so that the inventor can benefit from it is so completely against the grain of an enlightened world that it hardly needs mentioning - but there will be people reading this who will genuinely believe - just like the Victorians did about their life-long human captives - that Slavery is perfectly acceptable because "everyone does it".

• Banning of Articles of Incorporation with "profit" as the main focus. This is absolutely essential. There are plenty of alternative Articles of Incorporation where profits are made, but not at the expense of world resources. Companies House has a boilerplate designed for social clubs.

• FCC and other Spectrum Licensing. The process needs to be much more open: cooperation is the key, not competition. Not entirely sure how this should go, but it's mentioned here for completeness. Ultra-Wide-Band, with the means to operate under the noise threshold, in combination with ceramic phased-array antennae, pretty much makes it pointless to have licensing of spectrum, anyway, anywhere in the world.

• Other. I know I've missed a few things, here - I just don't know what they are.

Projects for a World-class Cooperative Economy

Here's a hint at the kind of projects that need to happen, which will give you an idea of why the above enormous list of hardware and software requirements is actually relevant. The list itself is also pretty big, yet the total resources of the Free Software community as measured in 2006 - two years ago - exceeded those of the world's largest corporate software house by 50%.

Bottom line: it's perfectly feasible. no - overachievable.

• Wikipedia

  Wikipedia needs to be turned into a distributed peer-to-peer application. This would actually be a fairly simple task, if the databases provided the object-orientated and distributed functionality described above.

  The data can easily be entrusted to a azureus-localhost, gnunetd file share, or freenet, with the servers that are currently being used to store the database ensuring that all of the data is always stored on the azureus-localhost, gnunetd or freenet distributed file share. Or, a gnunetd plugin could be written which provides a distributed front-end to a SQL database. a distributed SQL database.

  Ironically, this would give the Wikipedia Foundation exactly the kind of kick up the backside that they really need, because the distribution of the data would mean that they could be made entirely redundant. or replaced.

• Free Software replacement for Skype

  Skype is the only working Internet communications system, and it's not available for Linux-based smartphones, for example. Smartphones are incredibly complex bits of kit, as any embedded designer will tell you. Linux smartphones are extremely rare: before OpenMoko's "FreeRunner", you had very few choices: the HTC Universal Reverse-engineering effort has been going on for over three years, and the full hardware support (e.g. for things like switching the five speakers or the three microphones) is still not complete. There are also a couple of IPAQs (ironically also manufactured by HTC) such as the hw6915.

  Googletalk doesn't entirely cut it, but it's a good starting point.

• Language Translation Technology.

  Both computer languages and natural languages are contextual. There exists a plug-in for Visual Studio which performs translation of soruce code into *any* supported programming language. It's done by compiling down to CLR and then de-compiling (pretty-printing). This isn't rocket science. Computer Language translation is a perfect example of how the process of parsing which we as software developers take for granted needs to be automated.

• Automatic generation of Maps . Openstreetmap is a good start, however it's being carried out mostly by geeks who go war-driving, and they like to provide the locations of drinking establishments as a priority over-and-above things like speed cameras and hospitals. Uploading the routes taken by conscientious users who like to provide information other than drinking establishments would be extremely useful. And would avoid exorbitant licensing fees from government sanctioned organisations like the U.K. Royal Ordinance Survey. In addition, the same conscientious users might also like to add descriptive information about any other sights that they see along a route, such as pretty flowers, majestic views, cow dung and fly-tipping incidents.

• Replacing the proprietary A-GPS system As people walk around with their spangly-new linux-based hybrid device, it would be great to be able to record the signal strength of the cell-towers, triangulate their position, match that against the GPS coordinates, and upload into a distributed database the information derived. Then, when someone else switches on their GPS device, they can get the ID number of all the cell-towers in their area, obtain from the distributed database the approximate GPS coordinates, and feed that to the built-in GPS chipset so that it can get a quicker cold-start lock.

• Rsync with a VFS plugin layer would be an incredibly powerful system. Rsync itself is incredible, yet it is increasingly being focussed on hierarchical filesystems - Unix filesystems, at that. Rsync actually is good for synchronising data across hierarchical tree structures, yet it has been restricted to just Unix. a VFS plugin would allow people to back-end hierarchical data into TAR compressed archives; XML data structures; IMAP mailstores; and that's just the beginning. Note: Both FUSE and Samba's smbfs command-line utility demonstrate that it's not rocket-science to design a VFS layer, even in userspace. About 25 functions are required. smbfs basically over-rode the libc standard functions for __open, __close etc. - at runtime (LD_LIBRARY_PATH). Redhat 5 released a version of libc in 1998 which removed the required functions, effectively killing the project, but the principle is still there.

• IMAP with Rsync is an incredibly powerful combination, and in a peer-to-peer environment which does distributed store-and-forward messaging, with an SMTP front-end thrown in on top, even more so. The SPAM issues we face with email would entirely disappear overnight with such a powerful combination. And we would be able to sync email communications across many devices, by having the rsync-enabled IMAP server running on our own PDA, or Desktop, or anything. It's Blackberry on steroids [by the way, are people aware that the entire worldwide Blackberry infrastructure runs off of two servers - one in Canada and the other in the United Kingdom?]

• GIT needs more recognition, and for its back-end to be made a VFS plugin layer. GIT is a type of knowledge store, and it goes to a lot of trouble to "compress" the data and to optimise use of network traffic. GIT combined with rsync is an incredibly powerful combination: synchronisation of portions of a GIT repository, where the data doesn't have to be unpacked into a unix filesystem because rsync has a GIT-aware VFS plugin.

• l4linux.org needs to be mainstream. Only a year ago I heard of an HDTV embedded product being developed. It was a failure. The reason was that in an embedded environment, the latency of the Linux Kernel killed the whole system. Whilst, overall, the hardware spec looked "good enough", with the AVERAGE time looking adequate, the response time (for important things like pressing buttons on the front of the unit) was so dreadful that the whole project had to be canned. This is one area where Linux kernel developers, whose focus is primarily on "Desktops" and "Servers", simply don't understand why extreme latency matters. The solution is to adopt the l4linux.org source code as a compile-time option in the mainstream Linux kernel source tree, and all that implies. OSKit 1.0 had "issues" - the l4linux team solved them.

• "N of M" crypto. In 1999 I heard about a cryptographic algorithm which could be used for "cooperative" encryption and decryption. In some voodoo-magic way, a key could be shared across M parties, and with some even more obscure voodoo, only N of those M parties were needed to perform encryption, and, incredibly, a totally different set of N out of M parties could do decrypts. In combination with Trust Metrics and the Voice-Fingerprinting, it becomes possible to provide, rather than a "Public" Key Infrastructure, a "Distributed Grouping" Key Infrastructure. The usefulness of this combination cannot be underestimated: it is possible for two or more parties to sign binding contracts with their voices alone. The implications of that are quite startling: Banking, Justice and Arbitration services can be set up at will, on demand! These kinds of things are regularly part of the Science-Fiction I read, yet they are only possible when you have the combination of technology outlined here.

• Javascript PyPy "front-end" (e.g. for use in Firefox). This one's a bit obscure, but is an illustration of the type of improvements necessary - bear with me. Javascript is a dog. Everyone keeps writing their own javascript interpreter: there is one for every single browser out there, and they are all incompatible and they are all too slow. The PyPy project is an on-demand compiler which has front-end and back-end technology: CLR (Common Language Runtime) is one of the back-ends. On the front-ends, python is the main focus, although I had heard of someone adding a "B" front-end, just as an experiment. So, on the face of it, pypy looks like a python .NET compiler - it most definitely is *not* just a python .NET compiler. The point is: by providing a Javascript front-end, it should be possible to plug pypy into browsers - including lynx and dillo (!) via simple pre-processing. The end-result is that you would have an independent "javascript" engine for browsers, which could be used by any free software project. Also, google's search engine would come out of the cold from the Web 2.0 phenomenon, because servers could be dedicated to running the AJAX code, creating the HTML and then indexing that. In this way, Web 2.0 AJAX sites end up being brought into the fold. (btw - it doesn't have to be google that does that - it could be anyone).

• Pyjamas and other technology like it needs a much higher priority. Pyjamas is a port of Google Web Kit to python. The principle behind these technologies is to turn Browsers into Desktop Applications. However, the use of this strategy actually has an interesting side-effect: it gets people to partition their applications correctly. The exception to that rule *has* to be GoogleMail and YahooMail Web 2 apps, as it is crazy to expect people's little PDAs and embedded smart devices to have 512mb of RAM (hence the reason why the javascript engine idea above is so vital, because gmail and yahoomail are just the beginning of the Web 2.0 revolution). Ultimately, it should be possible to have the same source code run under GTK, KDE, Athena Widgets, Curses and Web Browsers, because the front-end is written in Pyjamas or GWT, the communication between the front-end and the back-end is done using JSON-RPC, and the back-end server doesn't necessarily run over the Internet, it runs on the local user's device. The user then gets to choose the level of interaction they want - depending on their resources - yet they still have the choice of running the application locally. In combination with peer-to-peer distributed infrastructure, built in to the service, the goal of redundancy, fault tolerance and usefulness is over-achieved.

• Distributed Distributions such as Debian. Debtorrent was created to take the load off of the Debian Mirrors, many of which are beginning to have to only do partial mirroring and are creaking under the load. The security updates are simply not adequately mirrored at all. The entire Debian distribution system needs to become properly distributed, from end-to-end. Debian's prevalent use of PGP key-signing puts them into a unique position of being able to entirely develop and publish the entire Debian Distribution using peer-to-peer distributed technology. Utilising the "old" methods - email and web technology - is clearly not working and is clearly not scaling.

  Also, the neat thing about the Debian Distribution system - dpkg - is that actually it need not be "programs" that are distributed with it: it could be Video archives, or music, or DNS zone files (all digitally signed). And the application front-ends to "install" software also already exist. The Debian Distribution system could easily be turned into a multimedia broadcasting system, where anyone in the world has the right to broadcast media (or anything else for that matter). "Democracy" Player would immediately become "useful" if it adopted peer-to-peer-enhanced dpkg infrastructure, providing exactly the technology that the BBC's iPlayer should not have spent £150 million of tax payer's money on in under four years to attempt to placate us with DRM-crippled single-platform rubbish.

• 3D virtual office Desktop . After looking forward enormously to KDE 4, to see what exciting new technology would be brought out, I was shocked to see something that looks like Vista and XP. This caused me to think about what would actually be "useful" to people. After having seen Compiz Fusion, it occurred to me that a 3D virtual office environment would be easily achievable. Analogies exist between every single concept in the computing world and a real office: a "filing cabinet"; a "telephone"; a "desk". Why are computer interfaces only modelling the desk "top"?? If someone wandered into a real office and found that their wall calendar, their 5 ft high filing cabinet, their potted plants and wall clock had all been shoved onto their desk they'd hand in their notice immediately.

  Once you have moved "Beyond the Desktop Metaphor" (look up the book of the same name) you have enormous flexibility and power to work as you see fit, rather than to have your life dictated to by a desk "top". For example, a virtual reality "filing cabinet" could expand out in 3D into a "dungeon" of enormous cavernous proportions, with the entire Library of Congress a small dot on the lantern-lit horizon.

  If you believe that this is Science Fiction and that the technology does not exist, today, run Compiz Fusion on an Intel Celeron M Ultra-low-voltage 600mhz CPU with an Intel Extreme 3D 855 Graphics chipset, and check the CPU usage. You will be stunned to find that the main CPU runs at only 400mhz, even on a 1280 x 1024 screen.

  Of course - tie-in into gaming systems such as Second Life and Worldforge is the next logical step, and it's in gaming environments where the need for peer-to-peer technology is most definitely felt.

  Then you have a proper collaborative working environment with which people feel much more comfortable.

• Collaborative Working tools are needed in a collaborative working environment. Web 2.0 technology such as Google Docs just doesn't cut it - you need technology like Abi-Collab (abiword with collaborative editing). Imagine being in your 3D virtual office, talking to someone on a VoIP phone (cue bakerlite 3D rendition of a phone), you then want to show them a document, so you "virtually invite" them into your "virtual office". They are then granted access to the document (over a peer-to-peer distribution system). You and their avatar in your "virtual office" begin discussing the document, and you invite your friend to make his own changes....

  As the ODF XML-based document is being updated by each user, rsync with an ODF-aware XML-based VFS plugin is being used to synchronise the changes to each other user...

• Auto-RDF, Semantic Web etc. takes a little explaining, but here we go: using reverse-engineering techniques, accelerated by automatic-generation of parsers (e.g. the hachoir author's reverse.py) it's possible to create "information merge" technology. For example, it would be trivial, using auto-parse technology, to create a web site which merged all of your social networking Web 2.0 identities, friends etc. This technology does exist, but it is usually used to "import" from one Web 2 web site into another.

  (Note to government agency people who may be wondering what that's all about - look at it this way: all those disparate systems that you have? it's possible to "merge" communications between them, without having to ask for any information from the contractors who charged you an arm and a leg ten years ago to develop those creaking - and non-contractually-compliant - systems. how is this done? by using automated reverse-engineering to analyse the communications, it's possible to VERY quickly have *modern* code written - mostly automatically - that can interoperate with each system. once you have done that with all the systems you want to connect together, you can then communicate via a common framework between all the different systems. It's not difficult: you're just being told that it is, so that you can be charged more of taxpayers' money).

• Distributed Filesystems for automatic backups are regularly created but are not ubiquotous. Again, this comes down to the "Desktop" metaphor, to the fact that Linux implements filesystems in kernel-space rather than user-space, and to some extent to the fact that the Unix userid space is limited (unlike in VMS, where user identification uses approximately 128 bits rather than only 16 or 32).

  FUSE - Filesystem in Userspace - for Linux is a "hack" that many "purists" deem unacceptable, particularly because, due to the monolithic and limited design of the Linux kernel, critical filesystem structures are locked for userspace applications to make filesystem accesses, potentially resulting in total and unrecoverable deadlock of the Operating System. This kind of design is unacceptable, yet there really aren't any alternatives that have as much developer focus. The GNU/Hurd has so much to catch up on that its far superior design is not able to take hold.

  Distributed Filesystems - or Global Filesystems - would allow groups of people - friends - to mirror and back up each others' files, independent of a centralised server infrastructure. A group of Developers would be able to share files, help compile applications, when on the move, when separated by a few miles, or when separated by continents.

• Peer-to-peer distcc (and dist.net and distjava etc.) - imagine ccache and distcc combined. Now imagine distcc with a peer-to-peer distsributed database. The MD5 hashes on the compiler options can easily be used as the DHT "key" into the distributed database. The result is that even a tiny device or a very slow embedded processor should be able to "compile" a binary in record time, by being able to download object files from all over the world. "make -j200" should be absolutely commonplace.

  p2pdistcc has other advantages as well: it makes it possible to do away with binary distributions, but still have the advantage of "downloads" of "stable" binary distributions. One of the reasons for having "binary distributions" at all is because of the time it takes to build them: if the object files are regularly precompiled by developers and made available in a peer-to-peer database, then not only does collaborative development go faster, but also the distribution maintainers job is done easier and faster, with no additional load placed on the distribution web servers.

Why??

The goal is to lower the barrier to entry the means for people to uplift themselves. That means that they need to be able to express their needs and desires, and then find someone who can fulfil them.

To achieve that, both hardware and software need to be robust, resilient and interchangeable. One of the side-effects of that is that the hardware and software will be free from corporate and governmental control. If I had said this even one year ago, I doubt that it would have gone down too well. However, it is becoming increasingly clear that the status quo is going down the pan, taking us with it, unless we act. Whilst we still have a chance.

So, by combining peer-to-peer software over peer-to-peer-enabled hardware, we end up with a self-healing, self-organising network that is useful to its users, even when part of the network is cut off from the rest of the world.

If you've read Neal Stephenson's work, you will have immediately recognised "The Librarian". The technology behind that science fiction piece of software - ironically developed by the merging of the CIA and the Library of Congress - really, really exists. It's just ... all over the place. This article points out how that technology can be drawn together to create a truly, truly World Economy worth living in.