
February 2004
Addressing The Challenges Of Real-Time
Fax Over IP
BY JAMES P. RAFFERTY
During the past twenty years, businesses of all sizes have come to rely
on fax communications as a sure-fire way to connect with people and
businesses throughout the world. There is a myth that fax machines and fax
servers are in a steep decline, but this simply does not square with the
facts that have resulted in the installation of some 100 million fax
machines worldwide, plus a healthy population of fax servers that have
been very popular within enterprises. In practice, fax technology and
standards have continued to evolve to address expanding business needs for
document communications of all kinds. For example, in recent years, many
enterprises have migrated to use V.34 facsimile, which offers page
transfer rates that are often twice as fast as those supported by the
older V.17 fax modems.
One of the hottest trends for today�s enterprises is the move toward
IP telephony. IP capable PBX sales have reached the tipping point in 2003
and now exceed the sales of PBX models that do not support IP networks.
However, with new technology, there are still some familiar challenges.
Today�s business still typically has numerous fax machines and fax
servers that were designed to operate over analog or digital phone lines.
As a business migrates its voice support to run over an in-house IP
network, it is important for IT managers to devise a game plan to provide
the same level and quality of fax service over IP as was previously
provided over the traditional PSTN.
In the late 1990�s, the fax experts in the ITU and e-mail experts in the
IETF saw that communications over IP was an important emerging trend and
the two groups agreed to work together to devise standards for Fax over
IP. As a result, in 1998, two approaches for Fax over IP were
standardized. The IETF focused on extending e-mail in order to be able to
support fax effectively and produced a series of related RFCs 2301-2306,
which collectively defined a method sending fax images via e-mail. The ITU
endorsed this effort and produced a related standard called T.37, which
directly referenced the IETF standards. In parallel, the ITU-T developed a
standard approach for real-time facsimile communication over IP networks,
which resulted in the T.38 standard. As the migration to VoIP has become
widespread, T.38 has emerged as the key standard for ensuring that VoIP
capable networks can continue to support user requirements for real-time
fax communications.
Let�s consider the challenges that T.38 fax over IP has been designed to
address. Virtually every fax device in the world supports the T.30
standard for Group 3 facsimile, which was devised by the ITU in 1980 and
has continued to be extended in the years since. In 1980, the telephone
network was dominated by analog technology, so that T.30 was designed to
work effectively over even very poor quality links. Fax negotiations in
traditional T.30 operate at only 300 bits per second and have built-in
integrity and retry features to help the exchange of fax signals complete
successfully. Once the two devices have agreed on a common set of
parameters for the fax modem links and the fax page, they then �train�
on higher-speed modems such as V.17 and the fax page transfer will happen
after that at data rates up to 14.4 kilobits per second. This approach has
proven to be very robust over a variety of phone network conditions and
means that two T.30 compliant fax devices can connect and communicate over
phone links that may be analog, digital, fiber, or even satellite.
However, one of the challenges we have seen is that many of the installed
fax machines only loosely follow the T.30 standard. In particular, some of
the older fax devices may have numerous bugs in the way that they respond
to T.30 fax signals, especially during negotiations and training. As a
result, the most experienced makers of fax equipment such as fax boards
have fine-tuned their T.30 implementations to be somewhat tolerant of
errors that may occur during the negotiations and training phases of T.30.
These more forgiving implementations of T.30 are thus able to connect with
a much higher percentage of the worldwide fax installed base than devices
that are less tolerant of protocol or network problems. Over time, fax
implementers have developed a collection of techniques for �keeping
alive� fax sessions in the face of such problems. These techniques are
often referred to as Fax Spoofing.
When developing the T.38 standard, the expert group�s focus was being
able to reproduce the real-time facsimile experience for end users while
running over packet networks rather than the PSTN. In particular,
facsimile users like to be able to watch their documents transfer in real
time and then get a confirmation of a successful transmission immediately
at the end of the session. After considerable discussion, the group
decided that the new real-time protocol for Fax over IP should be able to
transparently support all of the same document communication capabilities
as T.30 and that there should be minimal elongation of the call as
compared to traditional fax. As a result, the new T.38 protocol utilizes
the same approach as Group 3 fax in its representation of document
characteristics, but is designed to run over standard Internet Protocol
transport layers such as UDP/IP and TCP/IP. Since the T.38 protocol
encapsulates T.30 messages, this enables gateway developers to quickly
transform T.30 signals and data into a functionally equivalent packetized
form for transport over the IP network. The T.38 packet sizes for
negotiation signals are very small and can be protected using a redundancy
method in order to guard against packet loss.
Operation of fax over packet networks offers some different challenges
than traditional fax. As a general rule of thumb, if a packet network is
sufficiently robust to be able to support VoIP effectively, a T.38
implementation should also work reasonably well. For example, reasonable
quality for VoIP typically requires a total round-trip time of about
250-300 milliseconds or less for each packet. The T.30 fax protocol has
timers that are measured in seconds rather than milliseconds. For example,
fax negotiation signals will be re-sent if no response is received within
three seconds. Therefore, even if there are occasional network delays of
up to one second, this still gives T.30 and T.38 devices sufficient time
to communicate without violating timeouts and causing the session to go
down. Consecutive packet loss is more likely to cause issues than delay,
especially during the fax negotiations phase, since a T.30 fax session
will disconnect if three consecutive signals are missing or unreadable.
The deployment cycle of T.38 has been somewhat delayed by a few factors.
First, there was a need to complete a related set of standards that would
control the session over the IP network. Initially this was done using
H.323 and later on, support was added by the ITU for SIP and Megaco. H.323
deployment for T.38 is common, but support for the use of SIP with T.38 is
still very new. In addition, the need for T.38 has been driven by the
deployment cycle of VoIP in the enterprise, which slowed greatly during
the telecom downturn, but has recently begun to pick up. Third, the
initial implementations of T.38 were designed as gateway implementations
and many of the companies were relatively inexperienced in communicating
with Group 3 fax devices. Fourth, since T.38 is a new technology, there is
a need for the suppliers of T.38 gateway technology to work together to
ensure that their products interoperate. Finally, we are just beginning to
see the first uses of native T.38 fax devices (also called Internet Aware
Fax devices) that can connect directly to an IP network without need for a
gateway.
As IT managers consider making the move to adopt T.38 fax, the following
factors should be taken into account when evaluating such a migration.
� Does the supplier of T.38 technology have experience in connecting
successfully to a high percentage of the existing Group 3 fax installed
base?
� Does the T.38 implementation interoperate properly with other T.38
implementations?
� Will you be able to minimize changes to existing software
infrastructure and applications when moving from traditional to IP fax
servers?
� Will the solution provide ways to eventually eliminate the need for
all current analog or digital fax lines in the targeted offices?
� Does the T.38 implementation work effectively for the use of fax in
your enterprise environment, including ability to connect with remote
offices, as well as outside firms such as customers, vendors, law firms,
and other outside services?
� Will T.38 be supportable on the existing internal LAN or will upgrades
be required to provide sufficient bandwidth for co-existence of voice,
fax, and data applications on the same network?
� Will the IP fax solution support T.38 in a native manner or require
the use of T.38 compatible gateways?
T.38 shows great promise as an effective solution for bringing facsimile
fully into the age of converged networks. Like Group 3 fax itself, a
successful T.38 implementation will both follow the standards and be able
to operate under less than perfect conditions. As noted, a network whose
packet loss and delay characteristics can properly support VoIP should
also prove sufficient for support of T.38 fax communications. As today�s
fax solution providers add IP fax support and the quality of the
underlying packet network also improves, the transition from traditional
fax to Fax over IP should become much smoother. As the move to IP PBXs
continues to accelerate, a wide variety of options for bringing T.38 IP
fax into your organization without disrupting existing business processes
should soon be available.
James Rafferty is senior product manager for Brooktrout Technology.
Brooktrout Technology is a leading supplier of media processing, network
interface, call control and signal processing products that enable the
development of applications, systems and services for both the New Network
(packet-based) and the traditional telephone (TDM) network. For more
information, please visit www.brooktrout.com.
If you are interested in purchasing reprints of
this article (in either print or HTML format), please visit Reprint Management Services online
or contact a representative via e-mail at [email protected] or by phone at
800-290-5460.
[ Return To The February
2004 Table Of Contents ] |