Public Data Network Access (c) CyberLirik
                          Up-To-date

þ What is a Public Data Network?
  Public Data Networks are networks that conform to the CCITT X.25
  specification. This specification covers all aspects of how
  users with terminals (ie. PC running Procomm or the like)
  connect with host computers that are attached to the network,
  or an interconnecting network (most Public Data Networks are linked,
  although you might have to pay a fee to cross a network boundary).

  Examples of some well known PDNs are:
  Tymnet (British Telcom)
  SprintNet (formerly Telenet; Sprint)
  AutoNet (Owned by ADP)
  Datapac (Canadian Govt.)
  Israenet (Israeli Govt.)
  ... and so forth.

þ Can PDNs support SLIP or PPP?

  PDNs are incompatible with SLIP, but it is possible to run PPP over them.
  However, such connections are relatively slow due to the end-to-end
  error correction protocols provided by X.25. These protocols are not
  only unnecessary with TCP/IP connections and point-to-point (PPP)
  protocols, they actually place severe limits on performance.
  Public Data Networks are most useful, with respect to the Internet,
  in that they can provide a connection to an Internet host emulating
  a dialup session. There are technical complications that make
  this more complicated than it sounds.

  So, PDN access supports PPP, but in general it will be too slow for,
  say, Web surfing.

  Use PPP through a X.25 PAD

  Does anybody have experience with "tunneling" PPP and SLIP through
  the PAD-service (X.29 over X.25)?

  What I want is to let people dial up their PAD-service and
  send their PPP/SLIP packets across the
  X.25 network into the PAD-login of my UNIX-machine.

  This should be possible, but I guess the
  PAD-parameter configuration is critical??

  Yes, that's of course possible, because that's the way I use PPP. Use
  the PAD parameters for the following settings:
  no escape character 1:0
  local echo off 2:0
  flow/control: RTS/CTS 5:2 (this is perhaps not a standard X.3
  parameter)

  PAD should not react on XON/XOFF signals 12:0
  Other important values might be 3:0 4:1 9:0 10:0 13:0 14:0 15:0.
  You need a PAD that supports CTS/RTS flow control, because I don't know
  about PPP software that supports
  XON/XOFF (although this would be possible with the right async map).


   Global Networks: X.25: Making the Classic
                                Connection

  It may be off the front pages, but X.25 remains a popular
                        technology in Asia-Pacific

Frame relay and ISDN are getting plenty of press, but in the Asia-Pacific, the tried-and-true is the
technology of choice. And that technology happens to be X.25. So maybe these days X.25 spawns
more yawns than headlines. But Asia-Pacific net managers have three big reasons for keeping it
in their sights: It's widely available, it's inexpensive, and it's reliable--not all of which can be said
about some of the flashier newcomers.Solid performance at a sensible price was enough for Ed
Binney, Australasia IT manager for Thomas Cook Group Ltd. (London). The travel and financial
services company was looking to link its branch offices in Australia and New Zealand. Binney
picked an X.25 service from SITA Group (London). "The two drivers were cost and reliability," he
says, "and we have some high-volume sites where it's already paying dividends."

But when does venerable give way to long in the tooth? X.25 can't run any faster than 64 kbit/s,
and its protocol-handling and error-checking procedures add considerably to network delay.
Some also say that the demise of X.25 is inevitable, citing as evidence its absence from the list of
tomorrow's data communications strategies.

Classic Characteristics

X.25 is an ITU-T standard that defines the interface between a packet-switching network and a
user's terminal equipment. It's nothing if not well-established, and it continues to defy the
naysayers. "For 10 years people have been telling me it's dead," says Richard Barnes, senior
product manager of OSI services at the SITA Group. "But it keeps on going."

Longevity means there's plenty of X.25 equipment on the market. There also are a lot of people
familiar with the technology, which lessens some of the administrative burden. On top of that,
X.25 uses switched virtual circuits, which means there's no need to reconfigure the network every
time there's a change.

When it comes to geographic coverage, X.25 services have it all over the new kids on the block
(only modem dial-up is more widely available). In many countries, frame relay isn't an option
(and might not be for some time), because the infrastructure can't handle circuits of greater than
64 kbit/s. X.25 makes relatively few demands in that regard. Simon Brown, alliance manager,
global data solutions, at Telstra Corp. Ltd. (Sydney, Australia), even reports witnessing a working
demo where X.25 packets were sent along a string treated with electrolyte. That's probably a
more primitive setup than Asia-Pac net managers are likely to encounter, but it illustrates the
point.

What's more, X.25 makes good economic sense. Small companies making their first foray into
international communications find it a good starter technology: They can take advantage of
widespread coverage to get low-cost access. Other companies have their own reasons. Cook's
Binney considered ISDN but found that costs would rise as traffic increased. "The problem with
ISDN isn't startup costs, but the potential for escalating costs," he says. Other users point out that
because X.25 can handle multiple virtual circuits over a single access link, it's much cheaper
than leased lines, which require a separate access link and interface.

X.25 also is reliable--much more so than modem dial-up--thanks in large part to comprehensive
error checking. This might add some overhead, but as Ross Goodfellow, business development
manager for Scitor ITS Pte. Ltd. (Sydney, Australia), points out, there's really no noticeable
impact on throughput until speeds of more than 32 kbit/s are desired.

All of this might explain X.25's continued popularity in the Asia-Pacific. "There's a strong
marketplace for X.25 in developing countries," says Marc Olivier, of Asia-Pacific marketing and
sales, British Telecommunications PLC (BT, London).

Choose a Carrier

Those are some of the reasons why X.25 remains a widely deployed technology. But actually
taking advantage of it means picking a provider, and that's where things can get dicey.

Because of the intricacies involved in negotiating with telcos in various countries (such as
ordering local tails, dealing with language barriers, and obtaining support), this story focuses
instead on the nine international providers offering X.25 services in Asia-Pacific (see Table 1).



Even then, there's a complex overlay of marketing arrangements, alliances, and joint ventures.
Some of those local carriers thus end up marketing the services of the international providers. For
example, X.25 offerings from Infonet Services Corp. (El Segundo, Calif.) are sold by Kokusai
Denshin Denwa Co. Ltd. (KDD, Tokyo), Singapore Telecom Ltd. (Singapore), Telstra, and others.
In some countries, access to services from Compuserve Inc. (Columbus, Ohio) is via Scitor. In
addition, some carriers offer services from more than one provider: Telstra, for instance, resells
services from Infonet and Global One Communications (Reston, Va.).

Other carriers veer slightly from offering X.25 alone. GEIS International Inc. (Rockville, Md.) is a
second-tier X.25 player, focusing more on providing EDI and messaging solutions rather than
raw X.25 capacity. IBM Global Network (Paris) is something of a nonstarter in the X.25 stakes,
but it has real strength in the SNA market (which is its focus). Its Asia-Pacific X.25 presence
consists of just two nodes in Japan.

Location, Location, Location

The single most critical issue to consider when choosing an X.25 provider is geographic coverage,
and this is what most strikingly differentiates the players. Offering the most comprehensive
coverage are BT, Global One, Infonet, and SITA. BT's service is supplied by Concert
Communications Co. (Reston, Va.), the alliance it formed with MCI Communications Corp.
(Washington, D.C.). Cable & Wireless PLC (London), Compuserve, Tele Media International Ltd.
(TMI, London), GEIS, and IBM have less extensive coverage (see Table 2).



International coverage is one thing, but domestic coverage also should be considered. In larger
countries it's vital for a service provider to have nodes in all major centers of population--otherwise
there could be hefty access charges.

Prospective customers also should look at how reliable the carrier network is. If the net is poorly
meshed, traffic can't be rerouted to alternate circuits in the event of failure. To get a sense of
redundancy, net managers should ask how many bearers the service provider has on each of the
routes planned for use. Another strategy? Ask outright for a network map.

Rusty Gateways

There's more to look into. Net managers should make sure that the coverage offered is provided on
the operator's own network--not via an X.75 gateway or other link to a public network. An X.75
gateway can create a bottleneck, and it inevitably means low performance, since it's a single
point of failure.

"X.75 gateways are the pits," says Scitor's Goodfellow. "There are problems with interoperability.
You can't trace faults end-to-end, and they slow down response time. The equipment on the other
side of the gateway is not under your control, so you don't know about the management or
maintenance of it."

How to avoid X.75 gateways? Check for comprehensive coverage--it typically translates into
one-stop shopping. A single contact means that the provider handles all local, international, and
third-country connections; there's also just one bill to pay.

Still, there is one potential drawback to the single-carrier strategy: fault-tracking. If the X.25
breakdown occurs in the local access network, customers could find themselves in the middle of a
wrangle between the service provider and the local telco. And odds are no one will own up to the
problem.

Nevertheless, Brits Australia Rentals Ltd. (Melbourne, Australia), a vehicle-for-hire company,
made its carrier choice based on the one-stop principle. The company has offices in Australia,
New Zealand, South Africa, and Germany, and it uses X.25 for branch connectivity to the WAN.
It chose Scitor. "It's saved us a lot of headaches," says Phil Degenhardt, MIS manager. "It's good to
have the whole thing managed so we don't have to worry about the tail circuits or regulatory
issues."

Into the Mix

While X.25 is a standardized technology, X.25 services can be carried over a network running
other protocols. Some operators run proprietary protocols on their networks, encapsulating traffic
or doing conversions prior to transmission.

Most providers use a platform that handles a variety of traffic, possibly including X.25, frame
relay, SNA, or TCP/IP. This also allows them to offer gateway functions among various networks.
For example, Infonet has a gateway between its X.25 network and its Infolan IP services, so IP
and X.25 clients can exchange data on both networks. Beware: Flexibility like this carries a time
penalty. Every protocol conversion adds to the round-trip delay, so throughput can suffer.

Telcos normally provide packet-switching services on switched virtual circuits (SVCs), which
means that the network dynamically establishes for each packet the route with the least delay.
This is in contrast to frame relay, in which permanent virtual circuits (PVCs) set up a dedicated
path that remains in place all the time. This makes X.25 the right choice for linking sites that
generate small amounts of data traffic. It's also useful when connecting large numbers of sites,
since network changes won't carry PVC reconfiguration in their wake.

Access Plans

X.25 users have a range of access options. These include dial-up to a public port (X.28) or
leased-line or dial-up (X.32) to a dedicated port. X.28 is generally applicable to individual PCs
equipped with modems. X.32 is used mostly for connecting LANs or for leased-line overflow. It
offers higher levels of security and error correction than X.25, and it supports multiple virtual
connections over the dial-up link.

The disadvantage of X.32 is that it's more complex and expensive. Also, it's not offered by every
carrier. In addition, most operators offer PSTN (public switched telephone network) and ISDN
dial-back up for dedicated lines. However, ISDN backup is available only in countries where the
local telco can support it.

Access speeds, of course, vary according to access type. But geographic location also will play a
part. Every operator stresses that not all types and speeds are available everywhere, and that
sometimes different speeds are available upon application.

What a customer wants in access depends on what's needed. Dedicated lines allow users to call
or be called for sending and receiving data. They also permit higher speeds (up to 64 kbit/s),
although this is still pretty slow compared with the 2 Mbit/s typical of truly high-speed
communications.

With dial-up access, a user can receive and collect data--but at lower data rates. Sometimes the
modem will negotiate the rate down further, leading to even slower response. But service providers
report that dial-up access is experiencing substantial growth rates and now represents the
fastest-growing sector of the X.25 market. The reason? Mobile users, who need cheap, reliable
access to corporate databases from their laptops.

In most cases, leased lines are cheaper than dial-up when total connect time exceeds four hours
per day. Conversely, dial-up is less expensive than leased lines when connect time averages less
than two hours a day. For that gray area in-between, it's necessary to monitor traffic and costs to
get the best option.

Playing and Paying

For heavier or frequent users, a dedicated access line is the norm. In a typical X.25 network, a lot
of small remote sites are connected to a host. This requires the host to have a dedicated
connection because it needs the bandwidth and duplex capacity to receive data and to respond to
the remote sites. At the same time, the remote sites can be handled by a dial-up connection.

So that customers don't have to pay more than is necessary, a flat rate for a dedicated line is
recommended. This could be especially helpful in an SNA environment, where there is constant
polling. "A fixed cost is better value than paying per kilobyte," says Mauro Roselli, network and
communications support manager at Thorn Asia Pacific (Sydney), which uses X.25 for a
100-branch retail operation in Australia and for links to New Zealand and the U.K. "A fixed cost
helps our budgeting. We have to know our costs."

X.25 customers also will have to learn the intricacies of tariffs. These run the gamut from simple
time- or volume-based schedules to more complex packages.

Global One, for example, has three options for dedicated access, all of which carry access and
port charges: a per-kilobyte option; a capped option, in which the meter stops running when a
certain money level has been reached; and a monthly flat rate. Capping and flat-rate options are
useful when there are high volumes of traffic, especially if part of that volume is generated by
polling rather than by bona fide data traffic.

But even in flat-rate tariffs there are some subtleties. They're determined according to whether
there's a one-to-one (point-to-point) connection, an all-to-one (star or hub-and-spoke) link, or an
all-to-all (fully meshed) hookup. In other words, customers can choose the tariff most appropriate
to their networks. Economies of scale and changes in traffic patterns will determine the move from
several one-to-one tariffs to an all-to-one, and, finally, an all-to-all.

Finally, some X.25 carriers offer tools that let customers look into the network (although, of
course, they can't make changes). This is useful for monitoring traffic patterns and network
performance. SITA and Infonet both offer such tools. Sitavision can measure round-trip delays
between any two customer sites. Infonet's Perspexion provides for the electronic delivery of
performance and activity reports. These monitoring tools also can be used to track progress on
fault fixing, check on billing information, and analyze network parameters.


Diane Heather is the Sydney bureau editor for Data Communications