Network Working Group                                           J. WoodsRequest for Comments: 1979                                 Proteon, Inc.Category: Informational                                      August 1996                          PPP Deflate ProtocolStatus of This Memo   This memo provides information for the Internet community.  This memo   does not specify an Internet standard of any kind.  Distribution of   this memo is unlimited.Abstract   The Point-to-Point Protocol (PPP) [1] provides a standard method for   transporting multi-protocol datagrams over point-to-point links.   The PPP Compression Control Protocol [2] provides a method to   negotiate and utilize compression protocols over PPP encapsulated   links.   This document describes the use of the PPP Deflate compression   protocol for compressing PPP encapsulated packets.Table of Contents     1.     Introduction ......................................    2        1.1       Licensing ...................................    2     2.     PPP Deflate Packets ...............................    3        2.1       Packet Format ...............................    6     3.     Configuration Option Format .......................    8     SECURITY CONSIDERATIONS ..................................    9     REFERENCES ...............................................    9     ACKNOWLEDGEMENTS .........................................    9     CHAIR'S ADDRESS ..........................................   10     AUTHOR'S ADDRESS .........................................   10Woods                        Informational                      [Page 1]RFC 1979                      PPP Deflate                    August 19961.  IntroductionThe 'deflate' compression format[3], as used by the PKZIP and gzipcompressors and as embodied in the freely and widely distributedzlib[4] library source code, has the following features:       - an apparently unencumbered encoding and compression         algorithm, with an open and publically-available         specification.       - low-overhead escape mechanism for incompressible data.  The         PPP Deflate specification offers options to reduce that         overhead further.       - heavily used for many years in networks, on modem and other         point-to-point links to transfer files for personal computers         and workstations.       - easily achieves 2:1 compression on the Calgary corpus[5]         using less than 64KBytes of memory on both sender and         receive.1.1.  Licensing   The zlib source is widely and freely available, subject to the   following copyright:      (C) 1995 Jean-Loup Gailly and Mark Adler       This software is provided 'as-is', without any express or implied       warranty.  In no event will the authors be held liable for any       damages arising from the use of this software.       Permission is granted to anyone to use this software for any       purpose, including commercial applications, and to alter it and       redistribute it freely, subject to the following restrictions:       1. The origin of this software must not be misrepresented; you          must not claim that you wrote the original software. If you          use this software in a product, an acknowledgment in the          product documentation would be appreciated but is not          required.       2. Altered source versions must be plainly marked as such, and          must not be misrepresented as being the original software.Woods                        Informational                      [Page 2]RFC 1979                      PPP Deflate                    August 1996       3. This notice may not be removed or altered from any source          distribution.       Jean-Loup Gailly        Mark Adler       gzip@prep.ai.mit.edu    madler@alumni.caltech.edu      If you use the zlib library in a product, we would appreciate      *not* receiving lengthy legal documents to sign. The sources are      provided for free but without warranty of any kind.  The library      has been entirely written by Jean-Loup Gailly and Mark Adler; it      does not include third-party code.   The deflate format and compression algorithm are based on Lempel-Ziv   LZ77 compression; extensive research has been done by the GNU Project   and the Portable Network Graphics working group supporting its patent   free status.2.  PPP Deflate Packets   Before any PPP Deflate packets may be communicated, PPP must reach   the Network-Layer Protocol phase, and the CCP Control Protocol must   reach the Opened state.   Exactly one PPP Deflate datagram is encapsulated in the PPP   Information field, where the PPP Protocol field contains 0xFD or   0xFB.  0xFD is used when the PPP multilink protocol is not used or   "above" multilink.  0xFB is used "below" multilink, to compress   independently on individual links of a multilink bundle.   The maximum length of the PPP Deflate datagram transmitted over a PPP   link is the same as the maximum length of the Information field of a   PPP encapsulated packet.   Only packets with PPP Protocol numbers in the range 0x0000 to 0x3FFF   and neither 0xFD nor 0xFB are compressed.  Other PPP packets are   always sent uncompressed.  Control packets are infrequent and should   not be compressed for robustness.   Padding      PPP Deflate packets require the previous negotiation of the Self-      Describing-Padding Configuration Option [6] if padding is added to      packets.  If no padding is added, than Self-Describing-Padding is      not required.Woods                        Informational                      [Page 3]RFC 1979                      PPP Deflate                    August 1996   Reliability and Sequencing      PPP Deflate requires the packets to be delivered in sequence.  It      relies on Reset-Request and Reset-Ack LCP packets or on      renegotiation of the Compression Control Protocol [2] to indicate      loss of synchronization between the transmitter and receiver.  The      LCP FCS detects corrupted packets and the normal mechanisms      discard them.  Missing or out of order packets are detected by the      sequence number in each packet.  The packet sequence number ought      to be checked before decoding the packet.      Instead of transmitting a Reset-Request packet when detecting a      sequence error, the receiver MAY momentarily force CCP to drop out      of the Opened state by transmitting a new CCP Configure-Request.      This method is more expensive than using Reset-Requests.      When the receiver first encounters an unexpected sequence number      it SHOULD send a Reset-Request LCP packet as defined in the      Compression Control Protocol.  When the transmitter sends the      Reset-Ack or when the receiver receives a Reset-ACK, they must      reset the sequence number to zero, clear the compression      dictionary, and resume sending and receiving compressed packets.      The receiver MUST discard all compressed packets after detecting      an error and until it receives a Reset-Ack.  This strategy can be      thought of as abandoning the transmission of one "file" and      starting the transmission of a new "file."      The transmitter must clear its compression history and respond      with a Reset-Ack each time it receives a Reset-Request, because it      cannot know if previous Reset-Acks reached the receiver.  The      receiver need not do anything to its history when it receives a      Reset-Ack, because the transmitter will simply not refer to any      prior history ('deflate' is a sliding-window compressor).      When the link is busy, one decompression error is usually followed      by several more before the Reset-Ack can be received.  It is      undesirable to transmit Reset-Requests more frequently than the      round-trip-time of the link, because redundant Reset-Requests      cause unnecessary compression dictionary clearing.  The receiver      MAY transmit an additional Reset-Request each time it receives a      compressed or uncompressed packet until it finally receives a      Reset-Ack, but the receiver ought not transmit another Reset-      Request until the Reset-Ack for the previous one is late.  The      receiver MUST transmit enough Reset-Request packets to ensure that      the transmitter receives at least one.  For example, the receiver      might choose to not transmit another Reset-Request until after one      second (or, of course, a Reset-Ack has been received and      decompression resumed).Woods                        Informational                      [Page 4]RFC 1979                      PPP Deflate                    August 1996   Data Expansion      'Deflate', as used in this standard, expands incompressible data      by approximately 14-18 bytes (8 bytes worst-case at the 'deflate'      level, two further bytes for the 'deflate' end-of-block and the      zero-length synchronization block header, two bytes of sequence      number, and two bytes to account for adding the PPP Protocol Field      to the transmitted data unit).      The BSD Compress draft proposal[7] describes an escape mechanism      for incompressible data that trades off a layering violation for      the irritating complications of variable and potentially      unpredictable effective MRU lengths.  That direct escape mechanism      (and much of the text of its description) is used here as well.      If an incompressible data packet does not fit within the MRU of      the link, the packet MUST be sent in its original form without CCP      encapsulation; PPP packets with significant data expansion that do      not exceed the MRU of the link SHOULD be sent in their original      form without CCP encapsulation.  In both of these cases, the      transmitter must increment the sequence number, as future      encapsulated packets will depend on the correct reception of some      number of unencapsulated packets.      When a PPP packet is received with PPP Protocol numbers in the      range 0x0000 to 0x3FFF, (except, of course, 0xFD and 0xFB) it is      assumed that the packet would have caused expansion.  The packet      is locally added to the compression history.  (Given the      definition of the 'deflate' format, a convenient method of doing      this is to locally "decompress" a stored-block header of the      appropriate length, followed by the actual data block; or the data      can simply be appended to the receiver's history, depending on      implementation details.)      Sending incompressible packets in their native encapsulation      avoids maximum transmission unit complications.  If uncompressed      packets could be larger than their native form, then it would be      necessary for the upper layers of an implementation to treat the      PPP link as if it had a smaller MTU, to ensure that compressed      incompressible packets are never larger than the negotiated PPP      MTU.      Using native encapsulation for incompressible packets complicates      the implementation.  The transmitter and the receiver must start      putting information into the compression dictionary starting with      the same packets, without relying upon seeing a compressed packet      for synchronization.  The first few packets after clearing the      dictionary are usually incompressible, and so are likely to sentWoods                        Informational                      [Page 5]