PPP over ISDN
This specification is intended for those implementations which desire
to use the PPP encapsulation over ISDN point-to-point links. PPP is
not designed for multi-point or multi-access environments.
"It is clear that there is never likely to be a single, monolithic,
worldwide ISDN."  The goal of this document is to describe a few
common implementations, chosen from the current wide variety of
alternatives, in an effort to promote interoperability.
Simpson [Page i]
Table of Contents
1. Introduction .......................................... 1
2. Physical Layer Requirements ........................... 1
3. Framing ............................................... 3
4. Out-of-Band signaling ................................. 4
5. Configuration Details ................................. 5
SECURITY CONSIDERATIONS ...................................... 5
REFERENCES ................................................... 5
ACKNOWLEDGEMENTS ............................................. 6
CHAIR'S ADDRESS .............................................. 6
AUTHOR'S ADDRESS ............................................. 6
Simpson [Page ii]
PPP was designed as a standard method of communicating over point-
to-point links. Initial deployment has been over short local lines,
leased lines, and plain-old-telephone-service (POTS) using modems.
As new packet services and higher speed lines are introduced, PPP is
easily deployed in these environments as well.
This specification is primarily concerned with the use of the PPP
encapsulation over ISDN links. Since the ISDN B-channel is by
definition a point-to-point circuit, PPP is well suited to use over
The ISDN Primary Rate Interface (PRI) may support many concurrent B-
channel links. The PPP LCP and NCP mechanisms are particularly
useful in this situation in reducing or eliminating hand
configuration, and facilitating ease of communication between diverse
The ISDN D-channel can also be used for sending PPP packets when
suitably framed, but is limited in bandwidth and often restricts
communication links to a local switch.
The terminology of ISDN can be confusing. Here is a simple graphical
representation of the points used in subsequent descriptions:
+-------+ +-------+ +-------+
R | | S | | T | | U
+---+ TA +--+--+ NT2 +--+--+ NT1 +---+
| | | | | |
+-------+ +-------+ +-------+
These elements are frequently combined into a single device.
2. Physical Layer Requirements
PPP treats ISDN channels as bit or octet oriented synchronous links.
These links MUST be full-duplex, but MAY be either dedicated or
PPP presents an octet interface to the physical layer. There is
no provision for sub-octets to be supplied or accepted. The octet
stream is applied primarily at the R or T reference points.
Simpson [Page 1]
PPP does not impose any restrictions regarding transmission rate,
other than that of the particular ISDN channel interface.
PPP does not require the use of control signals. When available,
using such signals can allow greater functionality and
performance. Implications are discussed in .
Control signals MAY be required by some of the framing techniques
described, and is outside the scope of this specification.
The definition of various encodings and scrambling is the
responsibility of the DTE/DCE equipment in use, and is outside the
scope of this specification.
While PPP will operate without regard to the underlying
representation of the bit stream, lack of standards for
transmission will hinder interoperability as surely as lack of
data link standards. The D-channel LAPD interface requires NRZ
encoding at the T reference point. Therefore, as a default, it is
recommended that NRZ be used over the B-channel interface at the T
reference point. This will allow frames to be easily exchanged
between the B and D channels.
When configuration of the encoding is allowed, NRZI is recommended
as an alternative in order to ensure a minimum ones density where
required over the clear B-channel, with caveats regarding FCS .
Historically, some implementations have used Inverted NRZ (merely
switching the sense of mark and space), in order to ensure a
minimum ones density with bit-synchronous HDLC. The use of
Inverted NRZ is deprecated.
Implementations which desire to interoperate with multiple
encodings MAY choose to detect those encodings automatically.
Automatic encoding detection is particularly important for
Primary Rate Interfaces, to avoid extensive pre-configuration.
Only simple encodings are currently distinguished.
The only reliable method of detection available is to switch
modes between the supported encodings. Transmission of the LCP
Simpson [Page 2]
Configure-Request SHOULD be tried twice for each mode before
switching in rotation. This ensures that sufficient time is
available for a response to arrive from the peer.
Max-Configure MUST be set such that the cumulative attempts
result in no more than 59 seconds of time before disconnect.
It is preferable that the usual limit of 30 seconds be
By prior configuration, PPP MAY also be used with other
encodings. Because of difficulty distinguishing them, it is
not recommended that these encodings be automatically detected.
Terminal adapters conforming to V.120  can be used as a
simple interface to workstations. Asynchronous HDLC framing
 is accepted at the R reference point. The terminal adapter
provides async-sync conversion. Multiple B-channels can be
used in parallel. Unfortunately, V.120 has a framing mode of
its own for rate adaptation, which is difficult to distinguish
from Frame Relay, and which can confuse in-band frame
detection. V.120 is not interoperable with bit-synchronous
links, since V.120 does not provide octet-stuffing to bit-
stuffing conversion. Therefore, V.120 is deprecated in favor
of more modern standards, such as "PPP in Frame Relay".
The "Bandwidth On Demand Interoperability Group" has defined a
proposal called BONDING. Multiple B-channels can be used in
parallel. BONDING has an initialization period of its own,
which might conflict with the simple detection technique
described above, and requires extensive individual
configuration in some current implementations when multiple B-
channels are involved. It is recommended that the PPP Multi-
Link Procedure be used instead of BONDING.
For B-channels, in the absence of prior configuration, the
implementation MUST first use bit-synchronous HDLC , as opposed to
other framings, for initial link establishment. This assumes that
circuit-switched communications are generally [host | router] to
[host | router].
By prior configuration, octet-synchronous HDLC  is recommended
where the network termination equipment interfaces directly to the T
Simpson [Page 3]
reference point, and octet boundaries are available at the time of
framing. Such equipment is likely to be highly integrated, and the
elimination of bit-synchronous hardware can reduce the part count,
resulting in lower cost interfaces and simpler configuration.
Octet-synchronous HDLC MUST be used with NRZ bit encoding.
For D-channels, by default no data service is expected. By prior
configuration, "PPP in X.25" or "PPP in Frame Relay" framing MAY be
Despite the fact that HDLC, LAPB, LAPD, and LAPF are nominally
distinguishable, multiple methods of framing SHOULD NOT be used
concurrently on the same ISDN channel. There is no requirement that
PPP recognize alternative framing techniques, or switch between
framing techniques without specific configuration.
4. Out-of-Band signaling
Experience has shown that the LLC Information Element is not reliably
transmitted end to end. The deployment of compatible switches is too
limited, and the subscription policies of the providers are too
diverse. Therefore, transmission of the LLC-IE SHOULD NOT be relied
upon for framing or encoding determination.
No LLC-IE values which pertain to PPP have been assigned. Any other
values which are received are not valid for PPP links, and can be
ignored for PPP service.
As an alternative administrative measure, multiple directory numbers
can point to the same physical access facility, by binding particular
services to each directory number. The called party identifier has
proven to be reliably provided by the local switch.
When a called party identifier is used, or when a future LLC-IE value
is assigned to PPP and the PPP value is received, if the LCP has not
had the administrative Open event, the call MUST be rejected.
Receivers MUST NOT accept an incoming call, only to close the circuit
or ignore packets from the circuit.
Simpson [Page 4]
5. Configuration Details
The LCP recommended sync configuration options apply to ISDN links.
The standard LCP sync configuration defaults apply to ISDN links.
The typical network feeding the link is likely to have a MRU of
either 1500, or 2048 or greater. To avoid fragmentation, the
Maximum-Transmission-Unit (MTU) at the network layer SHOULD NOT
exceed 1500, unless a peer MRU of 2048 or greater is specifically
Instead of a constant value for the Restart timer, the exponential
backoff method is recommended. The Restart Timer SHOULD be 250
milliseconds for the initial value, and 3 seconds for the final
Implementations that include persistent dialing features, such as
"demand dialing" or "redialing", SHOULD use mechanisms to limit their
persistence. Examples of such mechanisms include exponential
backoff, and discarding packet queues after failure to complete link
establishment. In some implementations, discarding the transmit
queue can temporarily remove the stimulus to retry the connection.
Security issues are not discussed in this memo.
 Simpson, W., Editor, "The Point-to-Point Protocol (PPP)", RFC
1548, Daydreamer, December 1993.
 Simpson, W., Editor, "PPP in HDLC Framing", RFC 1549,
Daydreamer, December 1993.
 Stallings, W, "ISDN and Broadband ISDN - 2nd ed", Macmillan,
 CCITT Recommendations I.465 and V.120, "Data Terminal Equipment
Communications over the Telephone Network with Provision for
Statistical Multiplexing", CCITT Blue Book, Volume VIII,
Fascicle VIII.1, 1988.
Simpson [Page 5]
This design was inspired by previous drafts of C. Frost, B. Gorsline,
D. Leifer, K. Muramaki, S. Sheldon, K. Sklower, and T. Sugawara.
Thanks to Oliver Korfmacher (NetCS) for European considerations, Dory
Leifer (University of Michigan) for technical details and called
party signalling, and Vernon Schryver (Silicon Graphics) regarding
handling of link misconfiguration and timeouts.
Special thanks to Morning Star Technologies for providing computing
resources and network access support for writing this specification.
Back to RFC index