--------------------------------BEGIN----------------------------------
Hayes Microcomputer Products, Inc.
705 Westech Drive
Norcross, Georgia
404/449-8791
8 June, 1988
===========================
OVERVIEW OF
CCITT RECOMMENDATION V.42
===========================
HISTORY
Work on the modem error control standard began with the
appointment of the Special Rapporteur on error control at the 1984
Plenary Assembly of the CCITT (International Consultative Committee on
Telegraphy and Telephony) in Malaga-Torremolinos, Spain. Meetings in
the United States on the subject began in early 1985, and continue even
now (and for the foreseeable future).
EIA TR30 is an Accredited National Standards Committee
operating under the authority of the American National Standards
Institute. It develops voluntary standards for the United States, and
proposes U.S. positions for consideration by Study Group D. There are
three subcommittees under TR30: TR30.1 studies modems, TR30.2 the
interface between DTEs and modems, and TR30.3 the interfaces between
modems and the phone network.
Because the CCITT is an agency of the United Nations (under the
International Telecommunications Union [ITU]), its voting members are
countries. Most countries are represented by their Postal, Telephone,
and Telegraph Administration (PTT), but the United States does not have
such a government agency. Instead, the U.S. CCITT National Committee,
operating under the Department of State, formulates and approves U.S.
positions in CCITT matters. It has five study groups, of which one,
Study Group D, is in charge of positions related to modems (input to
CCITT Study Group XVII) and data networks (input to CCITT Study Group
VII).
The CCITT is divided into fifteen groups by topic (some groups
have been eliminated but the numbering has not been changed). Study
Group XVIIs charter is to study Data Transmission over the Telephone
Network). The recommendations (standards) developed by Study Group
XVII are usually assigned numbers in the V series, such as the modem
standards V.22, V.22bis, and V.32; ISDN terminal adaption standards
V.110 and V.120; interface standards such as V.24; and error control
standards such as V.42.
The International Standards Organization (ISO) is made up of
the national standards-making bodies from each country (in the USA,
this is ANSI). In cooperation with the International Electrotechnical
Commission (IEC), Joint Technical Committee 1 develops Information
Processing standards. Subcommittee 6 develops standards related to
Data Communications, in particular the bottom four layers of the Open
Systems Interconnection (OSI) reference model (including the Physical
and Data Link layers). There is close liaison and cooperation between
SC 6 and Study Group XVII on many issues, including error control in
modems.
EXISTING STANDARDIZED ERROR-CONTROL PROTOCOLS
SDLC is included as a standardized protocol for two reasons: if
there is any such thing as a de facto standard, SDLC qualifies; and,
SDLC is the foundation for nearly all subsequent development of data
communications protocols. It pioneered such techniques as layered
protocols and bit-oriented transmission with frame check sequences,
zero-bit insertion for transparency, and flags.
The international standard version of SDLC is not actually a
protocol in itself, but a catalog or menu of elements and procedures
from which appropriate choices can be made to build an actual protocol.
It is made up of six major standards: ISO 3309 (basic framing), ISO
4335 (elements of procedure, or frame types), ISO 7809 (classes of
procedures, or groupings for various applications), ISO 7478 (multilink
procedures, for dividing a logical connection over several physical
connections), ISO 8885 (general purpose parameter negotiation), and ISO
8471 (address resolution procedures for switched environments).
X.25 is the primary protocol used to connect synchronous
computers to packet networks. It includes both the LAP and LAPB data
link layer protocols (LAPB, Link Access Procedure-Balanced, is a subset
of HDLC), and the packet layer (implementing multiple virtual
circuits). It was originally adopted in 1976, but has been enhanced in
all subsequent CCITT study periods (1980, 1984, 1988) to meet the
growing demands of users. Associated standards include X.75
(interconnection between packet networks) and ISO 7776 (standardization
of DTE implementations of X.25).
LAPB is the error control protocol used in the Hayes V-Series
System Products.
LAPD was developed in CCITT Study Group XI to serve as the
protocol for the D signalling channel on ISDN (Integrated Services
Digital Network) connections. It is an extended version of LAPB.
E-PAD (Sweden) received very little consideration. It uses
bisync-like framing (HDLC basic mode) and asynchronous transmission,
and was designed as a higher-level interface between personal computers
and X.3 PADs.
Tymnet developed X.PC as an attempt to provide X.25-like
functionality in asynchronous transmission environments. Unfortunately,
because of vagaries of the Tymnet network internals and other design
limitations, X.PC is actually far from an X.25 clone. Some early
support for X.PC in the standards committees evaporated when
performance differences were studied (X.PC, since it uses start-stop
async transmission, has more than 20% additional overhead above X.25
LAPB and other bit-synchronous protocols).
MNP (Microcom Networking Protocol) was the first attempt to
integrate a protocol into a modem for point-to-point error control.
The original MNP modems used async transmission, but later versions
have grown to include synchronous transmission and other features. It
is different from LAP protocols in many respects, including lack of
provision of the OSI data link layer service, lack of piggybacked
acknowledgements, no multiplexing, and combined acknowledgement and
negative acknowledgement (and busy) frames. The greatest concern from
modem vendors about using MNP has been Microcoms habit of releasing its
own products with extended features much in advance of releasing
specifications of those features to its licensees (who are also
competitors). For example, Microcom has recently announced products
incorporating classes 7, 8 and 9 of MNP, but has only released through
class 6 to licensees.
V.42 has been approved by Study Group XVII for forwarding to
the CCITT Plenary Assembly which will be held November 14-25, 1988, in
Melbourne, Australia. This is actually just a formality; there is
little or no chance of technical change to the recommendation at the
Plenary, and no political opposition is expected.
STRUCTURE OF V.42
V.42 can be used with full-duplex two-wire dial-up modems on
the switched telephone network that use asynchronous-to-synchronous
conversion. Applicability to other modem types (half-duplex, for
example) is for further study.
The protocol defined in the main body of V.42 is known as LAPM
Link Access Procedure for Modems. Its features are discussed below.
Annex A of Recommendation V.42 specifies an alternative
protocol which is claimed, by the parties submitting the specification,
to be compatible with classes 2-4 of MNP (V.42 does not contain the
term MNP). It has been included only for backward compatibility with a
portion of the installed base of error-control modems.
Backward compatibility features are added to standards for
transition purposes, and are often deleted from recommendations after a
reasonable transition period. They are segregated into annexes to
simplify this process.
V.42 specifies that a modem claiming full compliance to the
standard must implement all parts of the standard, including both the
primary and alternative protocols. Modems may implement a portion of
the standard and claim compatibility only (which is a very subjective
term).
Many exciting features and capabilities are planned to be added
to V.42, but all of these will aply to LAPM only, since it is the
primary protocol. The alternative protocol is provided for
compatibility with the installed base only, and none of that installed
base will have any of the enhancements in their standardized form.
Existing extensions to MNP are not standard, and never will be.
V.42 specifies only the internal aspects and interface between
modems, and some elements of the hardware DTE interface. It does not
specify any AT-command-like control capability. This is the subject of
work in progress in the USA (EIA TR30.2) and in the CCITT (Study Group
XVII, Question 13.4).
One of the few interface elements provided in V.42 is for flow
control on the DTE interface, required during times when the modem is
retransmitting data after an error occurs. V.24 circuit 133 (Ready for
Receiving) is used by the DTE to control data flow from the modem, and
will most likely be assigned as an alternative use of pin 4 (Request to
Send) in the ISO 2110/EIA 232-D connector. The flow control function
has been added as an integral capability of the Clear to Send signal.
The XON and XOFF capability is bilateral (can be used by either
the terminal or modem) and is similar to the CCITT X.3 usage.
Selecting XON/XOFF protocol, however, means that binary data flow is
not permitted because the user data may contain XON and XOFF characters
which would be misinterpreted as flow control. All Hayes products will
include the Transparent XON/XOFF capability (currently in the V-Series
products) of encoding data so that such imbedded control codes do not
interfere with the flow control processes.
LAPM FEATURES
A V.42 modem includes a capability such that the calling modem
will send a sequence of control-Q (XON) characters with alternating
parity (to limit the possibility of misinterpreting user keystrokes) to
the answerer. Field testing has shown this benign detection phase not
to cause interference (except in rare cases) with DTEs attached to
non-error-controlling answering modems. The answering modem will
respond with a sequence of characters (EC) which signals to the calling
modem that the answerer has V.42 capability.
The detection phase may be disabled in the originator, in which
case the protocol establishment begins immediately upon connection of
the call. This may be used in cases when the user is certain that the
answering modem has V.42 capability. Answering V.42 modems must be
able to handle incoming calls with the detection phaseenabled or
disabled.
LAPM is an extension of LAPB and LAPD. It uses basically the
same connection establishment and termination procedures, as well as
similar data transfer procedures. Implementors familiar with either of
these protocols should have no difficulty with a LAPM implementation.
The poll/final bit procedures allow one modem to force the
other to transmit a response. This improves error recovery
capabilities by bypassing timer expiration recovery mechanisms (MNP has
no way to force the other modem to transmit).
Providing separate frame types for ACK, NAK and Busy functions
improves protocol reliabilty and eliminates the chance of lockups due
to misinterpretation of frame contents (MNP uses a single frame type
for ACK, NAK, and Busy functions). V.42 includes an enhanced Reject
capability to improve error-recovery performance in the presence of
line noise: a REJ frame may be sent upon receipt of any errored frame
rather than waiting for subsequent receipt of a good frame. Modems
receiving a REJ frame requesting a frame not yet sent simply ignore the
frame.
The address field allows for differentiation of commands and
responses, and in the future will also allow for multiple simultaneous
virtual data paths between the modems (for remote configuration,
network management, or user data such as multiplexing multiple
terminals or other devices). It also preserves compatibility with
existing HDLC protocols and increases the likelihood of interworking
with other HDLC-based devices in the future (MNP does not use an
address field). Use of the address extension bit provides for
multi-octet addresses.
The large numbering base for information frames, provided by
Modulo-128 I-Frame sequence numbers, permits a larger window size
(number of outstanding frames) than would be permitted under modulo-8
sequence numbering. This improves performance on connections with long
propagation delays, such as satellite links.
The internationally standardized procedures using XID Frame
exchange for negotiation defined in ISO 8885 and used in both X.32 and
LAPD (Q.921) are also used in LAPM. Both standardized parameters (such
as the options listed below) and manufacturer-specific enhancements can
be negotiated through this mechanism.
Enhancements provided by particular manufacturers such as data
compression (in advance of standardization in this area) may be
negotiated through a mechanism defined in V.42 which is under
consideration for international standardization in ISO. It uses the
same formatting conventions as used for negotiation of standard
parameters and options, and allows manufacturers to use any character
string as an identifier for their defined parameters.
V.42 permits renegotiation of link parameters between the
stations at any time during the connection. This may be useful if line
or user data flow conditions change, and the modem (a particularly
intelligent implementation, obviously) determines that different data
link parameters would improve performance. This may occur, for
example, if the default window size of 15 proves to be insufficient on
a high-speed connection on a double satellite hop (which can be
determined by repeatedly reaching the window limit before
acknowledgements are received, forcing a wait).
Unnumbered information (UI) frames are used for break
signalling out-of-band with user data. Three types of breaks are
supported: in-sequence breaks which do not flush outstanding data,
expedited breaks that bypass user data, and destructive breaks that
flush all user data buffers waiting for transmission (in both
directions).
In some environments the length of the break sent is important.
For example, some systems use a short break (100 milliseconds or so) to
interrupt data flow, while a long break (1.6 seconds or so) is a
request for disconnection. LAPM preserves the break length (MNP does
not) up to a maximum of 2540 milliseconds in 10 millisecond increments.
In high-speed modems such as V.32, the scrambler used to ensure
a constantly-changing data pattern on the phone line (to keep the modem
clocks synchronized) uses an algorithm which has the effect of
spreading single-bit errors over more than 16 bits. Research has shown
that in some cases this scrambler algorithm interacts with the 16-bit
frame check sequence algorithm to produce a 50% probability that some
errors will be undetected by the frame check sequence. The V.42 32-bit
frame check sequence option eliminates this possibility because the FCS
algorithm encompasses more bits than the scrambler algorithm, detecting
all propagation of errors by the scrambler.
All V.42 modems must support 16-bit frame check sequence;
32-bit FCS is negotiated at connection time using XID frames and is
enabled if possessed by both modems (MNP does not have this
capability).
When using high-speeds on long propagation delay channels, a
large number of frames may be outstanding at any point in time. The
standard error recovery method in LAP protocols requires that if a
frame is received in error, it and all following frames must be
retransmitted (MNP uses this technique as well). Selective Reject
capability in V.42 allows only the frame(s) received in error to be
requested for retransmission; following frames need not be
retransmitted unless they were also received in error. Selective
Reject is also negotiated via XID at link connection time, and may be
used if enabled in both modems (the alternative protocol does not have
this capability).
V.42 modems may retransmit SREJ frames if it can be determined
from incoming I-frames that the requested frame has been sent by the
other modem but was once again errored. This improves performance on
high-error-rate lines.
It is sometimes desirable to run loopback tests to insure the
integrity of the data communications link. Recommendation V.54
specifies physical (modulation) loopbacks that can be performed, and
V.42 (in LAPM only, not MNP) provides for a loopback test of the error
control functions as well. It uses the standard HDLC TEST frames, and
may be enabled through XID negotiation.
FUTURE PLANS FOR LAPM
As has been previously noted, work is continuing at this time
in the USA and internationally to standardize additional capabilities
for V.42 modems. Some of these features are available in error-control
modems today in non-standard form, but many are unique to V.42 and not
provided as yet in any products. When these extensions are defined,
they will be provided as optional capabilities to preserve
compatibility with the installed base of V.42 modems, and will apply to
the primary protocol (LAPM) only. Even if techniques used in existing
MNP class 5 or greater modems were adopted entirely by the CCITT, these
would not be added to the V.42 alternative protocol because it is
frozen.
The similarity between the V.42 LAPM protocol and the LAPD-like
protocol used in the V.120 terminal adaption standard will permit the
development of rules for interworking between these devices. This will
allow devices on the ISDN to easily interwork with devices on the PSTN
(Public Switch Telephone Network) without significant protocol
conversion resources. Proper encoding of data at the ISDN terminal
adapter may even eliminate the need for modem pools at the ISDNPSTN
gateway.
DATA COMPRESSION
Certainly, data compression to improve throughput is one of the
most important issues in error-control work at this time, and a
standardized technique is likely to be approved through accellerated
procedures early in the next CCITT study period. Hayes and others have
already contributed documentation on existing techniques as a
foundation for this future work. More contributions are expected. The
result is likely to be an amalgamation of the best points of the
existing techniques.
DATA ENCRYPTION
This capability is still at the level of feasibility study.
There is some objection to doing this at the data link layer, with the
preference being providing security functions at higher layers such as
the presentation layer. Several issues, such as key management, have
yet to be addressed, and it is likely to be some time before this work
is significantly progressed.
ASYMMETRICAL AND HALF-DUPLEX OPERATION
Many existing error-control modems, such as the Hayes V-Series
Smartmodem 9600, use half-duplex or asymmetrical transmission
techniques to achieve high throughput at reduced cost. Most of these
modems use proprietary techniques (Hayes alone uses an international
standard protocol, LAPB, in its modem), and there is interest in
defining a capability to support these transmission methods in V.42.
Changes in the timers and acknowledgement rules may be necessary. The
study group simply ran out of time or this would have been included in
the 1988 version of V.42.
MODEM RATE NEGOTIATION (MULTI-SPEED MODEMS)
Although significant degradation of circuit quality during a
single call is quite rare, there may be some benefit to be gained by
the ability for the modems, based on error rates or other objective
factors, to request a change to alternative (slower) modulation methods
with improved performance (and to switch back if conditions improve).
These rules could also be used to select among various transmission
mechanisms at initial connection time if both modems have multiple
capabilities.
CHARACTER FORMAT INDICATION AND NEGOTIATION
Some confusion currently exists in error-control connections
due to the fact that the character format (parity, stop bits) is
independently set on each DTE-modem interface, with an 8-bit format
used between the modems. Rules are provided in V.42 for encoding of 5,
6, 7, and 8-bit data into protocol frames, but no method is provided to
coordinate this setting between the two modems. This may result in
unexpected or improperly formatted data being delivered in situations
where different settings are used. Existing non-error-control modems
have the same problem (they will fail to communicate if different
character sizes are set), but intelligent error-control modems ought to
be able to coordinate these settings and at least warn the user of
possible problems.
In some cases, however, it is desirable for the error-control
modems to pass along data with improper parity rather than cleaning it
up as done by current error-control modems. These include tandem modem
links in which part of the connection has error-control modems and part
does not.
TRANSPORT OF INTERFACE STATE INFORMATION
In addition to preserving user data, it is sometimes desirable
to have end-to-end carriage of interface state information. This may
occur, for example, if the remote device is a printer with a paper-out
signal that needs to be received by the host. V.120 has this
capability today, and a similar scheme could be added to V.42.
FORWARD ERROR CORRECTION (CELLULAR RADIO)
Cellular radio applications present monumental challenges to
modem designers. Not only do drop-outs occur during cell transitions,
but even normal traffic (the driving by of a large truck) can interfere
with the signal and produce significant fading and other impairments.
Error rates experienced can be as high as one bit in one hundred or
worse, which would cause any normal error-control protocol to break
down and not be able to transfer even a single frame (the human ear
masks the resulting noise, but a modem cannot). Forward error
correction, such as used in Compact Discs, could be applied to V.42
modems. Throughput performance might be halved, but half is better than
nothing.
STATISTICAL MULTIPLEXING (MULTIPORT)
As mentioned above in the address field discussion, the
capability exists in V.42 for multiple simultaneous virtual circuits
between the modems. This is often used in high-speed modems today to
provide for connection of multiple terminals or a terminal and a
printer at a remote site. It is desirable to be able to provide this
capability in an error-control modem as well.
NETWORK MANAGEMENT AND REMOTE CONFIGURATION
In large networks, there is a great need to receive status
reporting and diagnostic information from widely-dispersed equipment,
particularly at unmanned sites, and also to be able to set parameters
and run tests on these remote modems. Standards in the entire area of
OSI network management are under study in ISO and CCITT, and error
control modems are no exception. The multiple virtual circuit
capability of V.42 is an excellent way to perform out-of-band
management without interfering with user data flow. The goal in JTC
1/SC 21 and CCITT SG XVII Q. 9 is to accomodate heterogeneous
multi-vendor environments, interfacing with existing management systems
such as NetView.
MULTI-FRAME SELECTIVE REJECT
As described above, selective reject allows for retransmission
of only errored frames rather than all following frames. If several
frames are in error, a separate SREJ frame must be sent for each one.
In asymmetrical modems especially, in which the acknowledgement channel
may be running at only 1/48th the speed of the data channel, these SREJ
frames may take a long time to send, increasing the likelihood of
reaching the transmitter's window size and delaying transmissions
unnecessarily. A multi-frame selective reject capability will allow
several individual frames to be requested in one SREJ frame, thereby
reducing substantially the overhead on asymmetrical links.
HAYES SUPPORT OF V.42
Hayes believes strongly that V.42 is the error-control
technique of the future, consistent with existing standardized
techniques and independent of proprietary control. The achievement of
the goal of a standardized error-control technique will eliminate fear,
uncertainty, and doubt in the marketplace, greatly increasing the
demand for error-control capability. Being an international standard,
homologation of the same product into many countries should be greatly
simplified, and error-controll communication between countries
facilitated.
Until techniques are standardized in CCITT for such features as
data compression, proprietary techniques will be supported in V.42 via
the manufacturers option negotiation XID procedures.
Any V.42-based product obtained from Hayes will be able to
interwork with the installed base of Hayes V-Series modems which will
be upgraded to V.42 support.
New versions of Smartcom II and Smartcom III will be provided
which include parameter selections to control the V.42-related
enhancements to Hayes products. These programs would thus be
immediately usable with any Hayes V.42 modem.
Hayes has always had a strong commitment to standards, and will
continue to actively participate in the work on V.42 and other data
communications standards. This ensures that the interests of our
customers are taken into account in the standards-making process, and
also allows Hayes to respond quickly to include agreed enhancements in
our products.
Our experience in developing the V-Series System Products,
which use the international standard LAPB protocol, will allow us to
very quickly implement the similar LAPM protocol and provide these
capabilities both as upgrades to existing products and in new products.
---------------------------------END----------------------------------