Preamble
This document describes the original concept of the UPN.
The UPN Development Status document describes the current state of development. As development has progressed, an improved architecture has evolved which dramatically reduces the cost of most of the network nodes. The costings and architecture description in this original concept document are therefore obsolete.
A hypothetical publicly owned and operated computer network is described. This Ubiquitous Public Network (UPN) could support Internet-like services for a cost of about $200 p.a. to each node owner, regardless of connect time, number of simultaneous connections and quantity of data transferred. It would therefore be a relatively low-cost means for individuals and organisations to provide simultaneous, numerous and protracted data connections.
The fixed-cost nature of the UPN is an important advantage over alternative network infrastructures. Fixed-cost service encourages use, but more importantly, it encourages creativity.
There is little doubt that the UPN is technically feasible. Much of the technology necessary has already been developed and is in use in other networks and data links. Further development would be required to tailor these existing technologies to suit the UPN.
The legislative environment in Australia may be prove to be an impediment. However, if given knowledge of the potential advantages of such networks, it is imagined that the more forward-thinking of our politicians would make efforts to clear these obstructions.
Data communications between geographically separated locations generally involve at least one physical layer link provided by a third party organisation such as a telephone company. Those wishing to transfer data, be it to access the Internet, to send a fax, to provide information to others or for some other purpose are bound by the costs and constraints imposed by the provider of this physical layer link.
There may be reasons for an individual or organisation to consider alternative ways to transport their data. These reasons may include
This document describes a hypothetical publicly-owned network which would allow low-cost data transfer between any participating nodes. This Ubiquitous Public Network would be publicly owned in that the individually-owned nodes would comprise the only hardware necessary for the network, all the network specifications, interface specifications and software would be public domain, and access to the network would not be controlled by a governing body or organisation.
The purpose of this document is to outline (within the limitations of the author's expertise) design considerations for the UPN, to describe a possible form for the UPN, and to solicit comments on the concepts described. Comments from those with experience in networks (which would hopefully correct any misconceptions of the author) and from academic institutions who may wish to consider becoming involved would be particularly welcome.
If the comments received indicate that there is sufficient interest in the concept of the UPN and agreement on its viability, it is imagined that a UPN "Engineering Task Force" would be established. Further research, development and ultimately implementation would broadly follow the Internet model.
The idea of the UPN is based on existing networks and subsystems. The Internet itself is the prime inspiration. The Amateur Packet Radio Network includes TCP/IP-over-AX25 networks which have many of the characteristics of the UPN. TAPR, one of the most active groups involved in Amateur Packet Radio, is currently working on implementing a TAPR.ORG Intranet to begin to link together regional groups that want to be part of their network. TAPR is involved in spread-spectrum technology, and is making available (to members only) a 902-928 MHz 115 kbit/sec frequency-hopping transceiver for $399. The LATNET is a wireless Internet access system which provides an Internet on-ramp using commercially available spread-spectrum transceivers. The Cook Report's "Wireless As An Internet On-Ramp & Local Loop By Pass" surveys "The Technologies, The Players And The Prospects - Including An In Depth Look At Metricom". The European Radiocommunications Office site has much useful information on spectrum usage, and one of its pages prepared by the Low Power Radio Association discusses some of the problems relating to use of low power radio systems. Firlan and others supply infrared data link systems with ranges useful for the UPN. The Infrared Data Association (IrDA) has developed specifications for infrared data transmission across open space at speeds up to 4 Mbit/sec with a range up to one metre.
The author concluded from these and other sources that
When considering a novel system it is generally prudent to first search all available sources which may describe similar systems. If no very similar systems are discovered, it is generally then prudent to ask why this is so. The reason is generally one of the following:
Considering these possibilities in turn:
This leaves the likely conclusion that the lack of profit for the initiator is the reason a UPN-like system has never been developed.
The UPN will not be viable if it contravenes legislation or if legislation imposes fees or unreasonable conditions.
New telecommunications legislation will come into effect in July 1997 which reduces the barriers to provision and use of telecommunications. These changes are likely to reduce any legislative difficulties for users of the UPN. The detailed effect of the new legislation is difficult to predict, partly because a publicly owned network does not seem to have been explicitly considered and partly because future Ministerial Rulings (on which much depends) are unpredictable. Liberal Party Policy Statements give some indication of the likely consequences. Many relevent documents are available on the Department of Communications and the Arts WWW site, including papers titled Communications Futures Project. and Networking Australia's Future. Gilbert and Tobin provide a summary of the situation, present and future.
AUSTEL is the Australian Government body responsible for supervising the introduction of competition to Australian telecommunications. The author has asked (6 Jan 1997) AUSTEL to comment on the likely effect of the post-July 1997 regulations on networks like the UPN.
The author has reviewed the Telecommunications Act 1996 and concludes that while the legislation will effectively prevent use of some of the more efficient forms for the physical layer, a carefully designed UPN will probably not be subjected to significant regulatory interference or impediment.
In January 1997 the FCC in the USA granted 300 MHz of spectrum at 5.15-5.35 GHz and 5.725-5.825 GHz to the Unlicensed National Information Infrastructure Band (U-NII). This was a result of an initiative by Apple. The radio spectrum allocation is accessible by equipment from any supplier and is available to anyone, without licensing or air-time charges. It allows fast, reliable links without legislative costs or interference. Unfortunately, this recognition of the public interest is yet to be demonstrated by the Australian government.
To be of value to its users the UPN must provide a service the user requires at a cost lower than the alternatives.
For the purposes of this document, data communications streams are considered according to both the bandwidth requirements and the latency tolerance of the associated applications. Data streams are accordingly categorised as shown in the following table.
|
Data Stream
Class |
Application Type |
Bandwidth
Requirements |
Latency
Tolerance |
| 1 | Full-motion video, eg. movies, graphics-based games | High | Low |
| 2 | Real time audio, eg. telephone | Medium | Low |
| 3 | Static interactive graphics, eg. WWW graphics pages | Medium | Medium |
| 4 | Interactive text, eg. WWW text pages | Low | Medium |
| 5 | Non-interactive graphics, eg. fax | Low | High |
| 6 | Non-interactive text, eg. e-mail | Low | High |
It is suggested (without defensible justification, but with a knowledge of the author's use of data transfer services and of the additional complication of providing low-latency service) that the first version of the UPN would need to support classes 3 to 6 inclusive to be considered worth investigating by potential users.
The data security concerns so frequently expressed in relation to the Internet would all apply to the UPN. The public access nature of the UPN would increase the number of people who could potentially monitor or alter data streams, although this has no implication for the measures necessary to avoid problems.
Public domain dual-key encryption and decryption algorithms could be used where necessary for UPN data transfers, for example between an Internet user and his ISP. These algorithms seem likely to allow secure communication until such time as a working quantum computer is produced.
Encryption and decryption should be performed in the end node itself rather than in the user's computer, so that the node can appear to the user's computer to be an AT-command modem.
The UPN would be subject to attacks from those who feel it conflicts with their interests, and from those of the "destructive hacker" variety. This must be considered in every aspect of the design of the UPN. Both regulatory and technical attacks are likely. Both will probably be merciless and executed by a perpetual stream of people with a high level of skill and perseverance, and in some cases considerable influence over government.
Whether the UPN thrives or dies will be determined largely by its ability to survive these frequent onslaughts, and its robustness in the face of unrelenting adversity.
It is likely that those who feel that the UPN conflicts with their profit interests will attempt to limit its usefulness.
An article in "The Age" 17/12/96, page D3 titled "Carriers sink plan for neighborly link" describes an embryonic non-profit neighborhood LAN, and also says "According to regulator Austel, the plan to install cables outside Gunter's own property cuts across the Telstra/Optus duopoly, and their agreement is needed for Austel to issue a permit........ Optus said no to Gunter's proposal ........ Telstra said that the request was still being considered but added that approvals were rare".
Other types of regulatory attack may include attempts to
It may be that the best way to avoid vulnerability to these sort of attacks is to
Eventually the usefulness of the UPN to the public would make it very much less likely that these sort of attacks would succeed, and it may well be that there would be effective pressure to improve the regulatory environment for the UPN, which could allow faster, cheaper links.
The UPN will be subject to technical attacks on a number of levels.
There will be attempts to disrupt data flow at the physical level both with high-powered noise and with data. The most effective defense against this form of attack may be highly directional receiving antennas, and use of a part of the spectrum which is limited to line-of-sight.
There will be attempts to crash the individual nodes of the network by corrupting data or software. There will be attempts to crash the entire network, or disrupt its operation by nodes using false addresses or acting as "black holes" to data. No doubt many other forms of attack are likely. Defence against these sorts of attacks must be considered as a part of the node and network design.
A peer-to-peer mesh seems to be the only practical topology for an unregulated network with short hop distances in relation to total path lengths. An inherent advantage of a mesh is that its bandwidth increases as the number of nodes increases.
All routing schemes known to the author depend on each node having a unique address. This poses a number of problems in an unregulated network, where there is no entity responsible for allocating addresses. Nodes are likely to appear and disappear frequently.
There seem to be a number of options available for address allocation.
- Address allocation could "ride on the back" of some existing scheme. Each node could require an Ethernet card, which has its own unique address as a result of an existing infrastructure.
- A node's address could be the complete (ie. country code included) telephone number of the node owner.
- A node's address could be derived from its geographic location. This has the advantage of providing an element of information useful in some routing schemes, but has the disadvantage of potential duplication.
A method of encouraging users to establish reliable nodes with a high uptime and good range would be desirable, probably essential. Without such a method in place, parasite nodes would become common, which were only accessible while they had data to transmit, or which had only one physical link to the network.
Each node could prioritise its handling and retransmission of incoming packets on the basis of a function along the lines of
(number of packets the requesting node has handled successfully for me) x (hop distance to that node) / (number of packets I have handled for the requesting node).
Prioritising on the basis of this (or more probably a more complicated function) could "reward" nodes which contributed more to the network, by providing them with faster throughput and shorter latency. It would also result in performance penalties for nodes which requested a disproportionately large share of bandwidth for the service they provided.
Nodes will need to be able to detect errant behaviour (both intentional and malfunction-caused) by other nodes, and reduce the service they offer to these. This may mean that some limited, optional form of source routing is necessary, so a node can influence the path of packets in order to identify the errant node.
Considerable care would be required in the design of the prioritising function to prevent it having unreasonable effects, and to prevent it being vulnerable to having its effect manipulated by selfish net users.
This design is intended show what the author sees as possible, and also to inspire "that won't work well, but if you do it his way ..." suggestions from those more skilled in the various fields. Another intent of this preliminary design is to minimise the creation of new methods and protocols so as to hasten the initial implementation, but to allow for improved protocols, methods and hardware.
This design shows what seems possible using relatively inexpensive technology, and estimates initial network performance from this. It is imagined that after a time a UPN would come to include a variety of technologies with various cost, speed and complexity tradeoffs. This variety would probably improve network performance.
For the reasons given in the section "Network Protection and Security", it is suggested that the red portion of the visible spectrum around 660 nanometers is used, but that allowance be made for future use of other types of physical layer.
It is known ( Firlan ) that products are commercially available which use infra red to communicate outdoors at 10 Mbits/sec over a range of 450 metres. It is assumed that this product uses a laser source, that the receiver section incorporates a directional antenna and that all "reasonable" design measures have been taken with this product to maximise the range and throughput.
Oatley Electronics supply a cheap ($60 with laser source, $33 with IR source) and simple "Lasercomms" kit (#K19) which they claim operates at a range in excess of 200 metres. The author has modified one of these to run reliably at 57.6 kbits/sec (and somewhat unreliably at 115.2 kbits/sec) using a single LED as the IR source and a photodiode backed by a 150 mm diameter reflector as the sensor. This arrangement has a maximum range of about 18 metres. It is probably reasonable to assume that range would be proportional to the square root of the number of LEDs used in the source. A similar link with 100 LEDs as the source may therefore have a range of up to 180 metres.
IrDA transceivers (ie. the IR and analog subsystems) and ENDECs (encoder/decoders which interface the transceiver to a 16550 chip or serial port) are available from Hewlett Packard (HSDL-1000 and HSDL-7000), Temic (TFDS3000, TOIM3000 and TOIM3232) and others for typically $5 to $10 for a 115.2 kbit/sec set. Texas Instruments makes the TIR1000 ENDEC. Crystal Semiconductors make the CS8130 transceiver/ENDEC, which requires an external LED and photodiode. Elekon makes an IrDA receiver which uses an external photodiode. Some of the transceivers have provision for additional external LEDs. The author is currently testing arrangements of some of these devices in conjunction with lenses, reflectors and LED arrays to increase the range.
Motorola's Application Note AN1016 titled "Infrared Sensing and Data Transmission Fundamentals" includes the circuit of an IR receiver front-end amplifier which is said in their "Optoelectronics Device Data" handbook to be suitable "for range of up to 100 metres". Other ANs in this handbook also have much useful information on IR communications.
Consumer products which use infra red remote controls typically use two IR LEDs as sources and low gain receiving antennas, and operate at ranges up to about 5 metres. The data rates of these remote controls are generally only a few Kbits/sec.
Based on the preceding information, and assuming that performance using red (660 nM) light would be similar, the following will be assumed to be achievable physical layer performance characteristics using 1996 technology:
| Type |
Cost/physical link |
Source System |
Sensor System |
Range (metres) |
Data rate (Kbits/sec) |
| 1 |
High (expensive
components) |
Laser, dedicated to a
single physical link |
Photodiode with precision optics,
dedicated to a single physical link |
450 | 10000 |
| 2 |
Medium (cheaper
components) |
Laser, dedicated to a
single physical link |
Photodiode with reflector,
dedicated to a single physical link |
300 | 115 |
| 3 | Medium |
100 LEDs, 22.5 degree
3 dB half angle, or fewer LEDS with reflector |
Photodiode with highly directional
reflector, dedicated to a single physical link |
150 | 115 |
| 4 | Medium |
100 LEDs, 22.5 degree
3 dB half angle |
Photodiode with somewhat directional
reflector, 22.5 degree 3dB half angle |
50 | 115 |
Nodes must on average have more than two links for a useful mesh. Let us assume that each node is designed with capability for three external links plus a connection to the node owner's computer running simultaneously.
Consider a mesh network of nodes using type 3 links. The area reachable by each node is 150 x 150 x pi = 71,000 square metres. If each node has three links to other nodes then there must be three other nodes within this area, so the minimum node density required is four per 71,000 square metres, ie. one per 17,000 square metres. In a suburb where a typical urban block is 1000 square metres, on average one building in seventeen must have a UPN node if each node is to link to three others using type 3 links. In Australia, approximately one household in twelve has a modem.
Consider a data transfer across a mesh network comprising type 2 and type 3 nodes. Assume
Number of hops = 10000 / 100 /.5 = 200
When there is no other network traffic, and assuming that each node in the chain begins retransmission of a received packet immediately it has received and confirmed the header checksum ("cut-through" routing), then the end-to-end latency is
25 x 200 x 8 / 115000 = 350 mSec.
and effective payload data transfer rate is approximately 90% of 115 kbits/sec, ie.103 kbits/sec.
There will be a number of factors which result in these figures not being met.
Because Type 3 and lower numbered types of link are dedicated in that they only ever receive data from one other node, there is never any data collision. Collision detection, avoidance and recovery are unnecessary. This is a considerable advantage over shared channel links.
Because of the large sliding window size, packet receipt acknowledgements will be returned to the sending node in time to avoid delaying transmission of subsequent packets. A large sliding window has disadvantages in that it means larger buffers in the nodes, and (depending on the recovery process) may increase error recovery time.
The main cause of additional delays in a real system will be that a node will already be sending data over a link over which its needs to send the incoming data. This will result in a delay ranging from minor to extreme, depending on network loading, node interconnection density, routing protocols and other factors.
Another minor source of delay and throughput reduction may be the need for a node to determine the availability of buffer space in a node to which it wishes to send a packet, prior to actually sending the packet. This sort of flow control may be necessary to ensure graceful degradation of the network under heavy loading, and also to allow nodes to exercise priority control and thereby support an Incentive Scheme, as described earlier.
Let us assume at his stage that these overheads and other factors will reduce throughput to 25% of the ideal figure calculated above, and increase latency by a factor of 4. This would mean we had a data transfer rate of approximately 25 kbits/sec, and an end-to-end latency of 1.4 seconds. These are probably acceptable for class 3, 4 and 5 applications.
Allowance should be made in the design of the UPN for other types of physical links. Longer distance microwave links may be set up by ISPs or others. Medium range links may be established using spread spectrum technology in the ISM bands, although Australia's vigorous regulation and relatively low allowable power levels (see AS 4268.2 - 1995) would initially be a disincentive in this country. Short range links (eg. coax to the neighbours of a node-owner) may be useful to some.
Potential users will only be motivated to set up UPN nodes if they can derive immediate advantage. The advent of the Internet and the associated ISPs provides an opportunity for the UPN to provide this instant gratification. The ISP the author has described the UPN concept to has shown considerable interest.
Consider first the possibility of one ISP setting up a UPN node. The few technically elite who both use and are within a single hop of that ISP may set up UPN nodes, but would have no incentive to configure those nodes to relay data from other nodes. In rare circumstances there may be two technically elite users of that ISP who are friends, geographically within one hop of each other, but only one of whom is within a single hop of the ISP. In this situation the UPN may stretch from the ISP out two hops. The ISP could encourage the expansion of the UPN out from his site by facilitating communications between his customers.
The best ISP for the point source node propagation model would probably be a University, because there there is likely to be an unusually high proportion of technically competent individuals with free Internet accounts, who have an interest in technology, who tend to live in a small geographic area, and who communicate with each other more than the average ISP customers. Other advantages of the University include thorough testing of the robustness and vulnerability of the UPN, along with a capacity to devise ways to fix the crashes they may be able to cause. A disadvantage of the University is the generally woeful state of students' finances and the corresponding disinclination to pay for a node, but this may be balanced to some extent by a disinclination to pay telephone call charges for their Internet sessions.
Another node propagation model involves two UPN-connected ISPs who have customers spread in the general area between them. In this case, a UPN node chain could be established generally joining the two ISPs with advantages to all, provided each ISP had a customer within one hop of the other ISP. Once this sort of node chain were established, node propagation extending each side of the chain would be likely. The ISPs could establish WWW-accessible data bases or other means to advertise unused node links and geographic locations on behalf of their customers, again to the benefit of all.
Many problems are imagined.
Would people be prepared to set up UPN hardware, and have the associated "boxes and bits" attached to their houses? It is beyond many peoples capability to erect their own TV antenna and point it (roughly!) towards the TV transmitter's antenna, so how would these people go trying to affix and align optical link hardware in cooperation with someone a hundred or more metres away? Not well probably. Only a very small proportion of the population would have the capability to be involved in the UPN unless commercial organisations found that there was viable a market for turnkey installations.
If a mesh of UPN nodes exists in an area, and each node has near-maximum range links in three directions, how does a prospective new user in that area persuade one or more of these existing nodes to provide him with a link? There appear to be more reasons why one of these existing node owners should refuse, rather than offer another link. A mesh with branches seems inevitable.
Each node would determine his own address on the basis of his geographic location. Some means will be necessary to assist in address derivation from latitude and longitude, grid reference or map reference.
A method will be needed to detect and resolve address duplications. This could include a node refusing to register a new direct physical link to an address previously recorded in its local routing table, or to an address more distant geographically than is possible for that type of physical link.
A four-byte address field would allow two bytes to be used for the X coordinate and two bytes for the Y coordinate. This would allow a network to cover a 130 km by 130 km area with 2 metre resolution. A larger address field to cover the entire earth with perhaps 1 metre resolution may be preferable.
Each node would be responsible for relaying received packets towards the destination address. A suitable workaround would need to be found for the "dead end street" problem which is a characteristic of coordinate-based routing in incomplete mesh networks . This workaround may include a two-level routing system in which nodes each maintain and use a routing table for nodes within an certain number of hops from them, but use a distance-to-destination based routing logic for other destinations, along with optional source routing to the extent of specifying one or more intermediate nodes.
Standard PC.
Linux or derivative, or KA9Q NOS or derivative. Both are free, and some of the varieties and subsystems available are richly endowed with functions required for a UPN node.
The node would appear to the node operator's computer to be a standard AT-command modem connected to a telephone line.
A UPN could:
The total design cost for the UPN is likely to be both substantial and difficult to predict. No one organisation or individual could control the UPN, and so could not gain greater benefit from it than any other user. Commercial organisations could generally not justify expenditure on UPN development. Therefore, the only way the design work will be effected is by unpaid volunteers. These volunteers may come from several groups, including
|
Item |
Estimated Cost |
|
Second hand 386 computer for use as node controller
with 4 off 16550 comms ports |
$300 |
|
(3 off )
source modulator, driver and source, comprising - 100 LEDs @ $0.5 - PCB @ $10 - Components @ $10 - Enclosure @ $20 |
$270 |
|
(3 off)
receiver, preamp, demodulator, comprising - Reflector @ $15 - PCB @ $10 - Components @ $10 - Enclosure shared with source |
$105 |
| Cabling & connectors | $80 |
|
Power Supply for subsystem
Assume power (12 V @ 2.8 A max) available from computer |
|
|
Software
Assume free |
|
| TOTAL | $755 |
|
Item |
Annual Cost |
|
Power, node controller continuously running
but associated VDU generally powered down, 70 watts average @ $0.12/kwh |
$74 |
| Repairs and maintenance | $50 |
| Cost of capital @ 10% | $76 |
| TOTAL | $200 |
The benefits which the UPN can provide depend on
- whether local phone call charges are based on time,
- whether the UPN is widespread, and
- the usage profile of the individual or organisation.
The following tables show estimates of the benefits of a variety of combinations of these factors. These tables are based on the assumption that
- Untimed local telephone calls are charges at $0.25 per call
- Timed local phone calls are charged at $0.25 on connect plus $0.02/minute after five minutes
- The annual rental for a telephone line is $172 p.a. for residential, $240 p.a. for commercial
|
User Profile |
Approx cost of alternative p.a. |
Benefit of UPN p.a. |
|
Individual who connects to the Internet or University
each day, and can tolerate his phone line not being available for other uses during these periods. |
$91 |
$91-$200
=$-109 |
|
Individual who connects to the Internet or University
each day, can tolerate his phone line not being available for other uses during these periods, and who shares a UPN node with three others in the same house or next door. |
$91 |
$91-$50
=$41 |
|
Individual who connects to the Internet or University
each day, and can not tolerate his phone line not being available for other uses during these periods, and so requires an additional phone line. |
$91 + $172
=$263 |
$263-$200
=$63 |
|
ISP with ten lines for customer access
(assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$2400 |
$2,400-$200
=$2,200 |
|
University with 100 lines for student access
(assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$24,000 |
$24,000-$2,000
=$22,000 |
|
User Profile |
Approx cost of alternative p.a. |
Benefit of UPN p.a. |
|
Individual who connects to the Internet or University
each day for one hour, and can tolerate his phone line not being available for other uses during these periods. |
$401 |
$401-$200
=$201 |
|
Individual who connects to the Internet or University
each day for one hour, can tolerate his phone line not being available for other uses during these periods, and who shares a UPN node with three others in the same house or next door. |
$401 |
$401-$50
=$351 |
|
Individual who connects to the Internet or University
each day for ten hours, and can tolerate his phone line not being available for other uses during these periods. |
$4,400 |
$4,400-$200
=$4,200 |
|
ISP with ten lines for customer access
(assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$2,400 |
$2,400-$200
=$2,200 |
|
University with 100 lines for student access
(assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$24,000 |
$24,000-$2,000
=$22,000 |
|
User Profile |
Approx cost of alternative p.a. |
Benefit of UPN p.a. |
|
Individual who connects to the Internet, educational
institute or workplace each day, and can tolerate his phone line not being available for other uses during these periods. |
$91 |
$91-$200
=$-109 |
|
Individual who connects to the Internet or educational
institute or workplace each day, can tolerate his phone line not being available for other uses during these periods, and who shares a UPN node with three others in the same house or next door. |
$91 |
$91-$50
=$41 |
|
Individual who connects to the Internet , educational
institute or workplace each day, and can not tolerate his phone line not being available for other uses during these periods, and so requires an additional phone line. |
$91 + $172
=$263 |
$263-$200
=$63 |
|
ISP or small business with ten lines for customer / employee
access (assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$2400 |
$2,400-$200
=$2,200 |
|
School, university or company with 100 lines for student /
employee access (assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$24,000 |
$24,000-$2,000
=$22,000 |
|
User Profile |
Approx cost of alternative p.a. |
Benefit of UPN p.a. |
|
Individual who connects to the Internet , educational
institute or workplace each day for one hour, and can tolerate his phone line not being available for other uses during these periods. |
$401 |
$401-$200
=$201 |
|
Individual who connects to the Internet, educational
institute or workplace each day for one hour, can tolerate his phone line not being available for other uses during these periods, and who shares a UPN node with three others in the same house or next door. |
$401 |
$401-$50
=$351 |
|
Individual who connects to the Internet, educational
institute or workplace each day for ten hours, and can tolerate his phone line not being available for other uses during these periods. |
$4,400 |
$4,400-$200
=$4,200 |
|
ISP or small business with ten lines for customer /
employee access (assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$2,400 |
$2,400-$200
=$2,200 |
|
School, university or company with 100 lines for
student / employee access (assuming one UPN node allows ten concurrent virtual connections at acceptable speed) |
$24,000 |
$24,000-$2,000
=$22,000 |
1. While local calls remain untimed, there is probably only incentive for an urban-dwelling individual Australian to use the UPN in place of the existing telephone system if that individual
- accesses the Internet frequently, or
- needs protracted connection to the Internet, his educational institution or his workplace and wishes to leave his telephone line available for other uses during these periods, or
- can reduce the cost of his node (e.g. by using an existing spare computer for the node controller), or
- has a philosophical or technical interest in the UPN.
However, at least one Australian telephone company has already spoken of timing local calls for data connections, and this seems almost inevitable as the number and duration of this type of call is increasing rapidly. The "Herald Sun" of 4/12/96 says in an article titled "Net takes over phones" that Telstra spokesman Steve Nason "belives it is only a matter of time before timed Internet calls were introduced. He said Telstra had not yet set up the technology to differentiate between modem calls and someone talking on the phone, but it would not take long."
2. When local calls are charged on a time basis, the UPN could provide services at lower cost than the alternatives for most organisations and individual data users.
3. Irrespective of whether local calls are charged on a time basis, the UPN could provide services at much lower cost than the alternatives for organisations such as ISPs, educational institutions and companies who need to support multiple simultaneous data links.
Solicit comments on the technical viability of the UPN from others, especially Universities, Amateur Packet Radio groups, ISPs and other organisations and persons involved in similar technology. (This is the purpose of this document.)
Create a REV B design outline based on these comments.
Invite Universities, ISPs and others who may express interest to become part of a preliminary working group to refine the concepts and divide the project into workable parts with definable interfaces.
Assign the various parts of the project to organisations and individuals who express interest in contributing to its development and who appear to have both the means and the will to successfully perform the work within an acceptable time frame.
Determine and establish a suitable initial test site, probably a University or ISP who has participated in the earlier phases and who can encourage a useful number of individuals to set up nodes.
Refine the UPN based on the experiences at the initial test site.
Develop specifications and interface definitions.
Publish (on the Internet) all design data, specifications and definitions for the UPN and its components.
Establish a UPN Engineering Task Force with duties similar to those of the IETF.
A city-wide publicly-owned data network is probably technically viable. It could be a relatively low-cost means for organisations to provide simultaneous data connections to many users, and for individuals and organisations who need many or protracted data connections or who wish to make information available to others. It would be attractive to more individual users in areas where local telephone call charges were based on connect time.
There are many technical and structural difficulties for which solutions would be needed, but none of these seem insoluble. The difficulty of achieving a "critical mass" of users would be considerably reduced by the involvement of universities or ISPs. The project would require the support and contributions from a large number of skilled volunteers, especially in the earlier stages.
Derek Weston, derekw@alphalink.com.au
Melbourne, Australia
Thanks to VICNET for hosting the UPN Web Pages
hits since 7 Jan 1997.