💾 Archived View for gemini.spam.works › mirrors › textfiles › programming › zmodem6.txt captured on 2020-10-31 at 14:36:40.
-=-=-=-=-=-=-
The ZMODEM Inter Application File Transfer Protocol Chuck Forsberg Omen Technology Inc A overview of this document is available as ZMODEM.OV (in ZMDMOV.ARC) Omen Technology Incorporated The High Reliability Software 17505-V Northwest Sauvie Island Road Portland Oregon 97231 VOICE: 503-621-3406 :VOICE Modem: 503-621-3746 Speed 1200,2400,19200(Telebit PEP) Compuserve:70007,2304 GEnie:CAF UUCP: ...!tektronix!reed!omen!caf Chapter 0 Rev 10-27-87 Typeset 10-27-87 1 Chapter 0 ZMODEM Protocol 2 1. INTENDED AUDIENCE This document is intended for telecommunications managers, systems programmers, and others who choose and implement asynchronous file transfer protocols over dial-up networks and related environments. 2. WHY DEVELOP ZMODEM? Since its development half a decade ago, the Ward Christensen MODEM protocol has enabled a wide variety of computer systems to interchange data. There is hardly a communications program that doesn't at least claim to support this protocol, now called XMODEM. Advances in computing, modems and networking have spread the XMODEM protocol far beyond the micro to micro environment for which it was designed. These application have exposed some weaknesses: o The awkward user interface is suitable for computer hobbyists. Multiple commands must be keyboarded to transfer each file. o Since commands must be given to both programs, simple menu selections are not possible. o The short block length causes throughput to suffer when used with timesharing systems, packet switched networks, satellite circuits, and buffered (error correcting) modems. o The 8 bit checksum and unprotected supervison allow undetected errors and disrupted file transfers. o Only one file can be sent per command. The file name has to be given twice, first to the sending program and then again to the receiving program. o The transmitted file accumulates as many as 127 bytes of garbage. o The modification date and other file attributes are lost. o XMODEM requires complete 8 bit transparency, all 256 codes. XMODEM will not operate over some networks that use ASCII flow control or escape codes. Setting network transparency disables important control functions for the duration of the call. A number of other protocols have been developed over the years, but none have proven satisfactory. o Lack of public domain documentation and example programs have kept proprietary protocols such as Relay, Blast, and others tightly bound to the fortunes of their suppliers. These protocols have not benefited from public scrutiny of their design features. Chapter 2 Rev 10-27-87 Typeset 10-27-87 2 Chapter 2 ZMODEM Protocol 3 o Link level protocols such as X.25, X.PC, and MNP do not manage application to application file transfers. o Link Level protocols do not eliminate end-to-end errors. Interfaces between error-free networks are not necessarily error-free. Sometimes, error-free networks aren't. o The Kermit protocol was developed to allow file transfers in environments hostile to XMODEM. The performance compromises necessary to accommodate traditional mainframe environments limit Kermit's efficiency. Even with completely transparent channels, Kermit control character quoting limits the efficiency of binary file transfers to about 75 per cent.[1] A number of submodes are used in various Kermit programs, including different methods of transferring binary files. Two Kermit programs will mysteriously fail to operate with each other if the user has not correctly specified these submodes. Kermit Sliding Windows ("SuperKermit") improves throughput over networks at the cost of increased complexity. SuperKermit requires full duplex communications and the ability to check for the presence of characters in the input queue, precluding its implementation on some operating systems. SuperKermit state transitions are encoded in a special language "wart" which requires a C compiler. SuperKermit sends an ACK packet for each data packet of 96 bytes (fewer if control characters are present). This reduces throughput on high speed modems, from 1350 to 177 characters per second in one test. A number of extensions to the XMODEM protocol have been made to improve performance and (in some cases) the user interface. They provide useful improvements in some applications but not in others. XMODEM's unprotected control messages compromise their reliability. Complex proprietary techniques such as Cybernetic Data Recovery(TM)[2] improve reliability, but are not universally available. Some of the XMODEM mutant protocols have significant design flaws of their own. o XMODEM-k uses 1024 byte blocks to reduce the overhead from transmission delays by 87 per cent compared to XMODEM, but network delays still __________ 1. Some Kermit programs support run length encoding. 2. Unique to DSZ, ZCOMM, Professional-YAM and PowerCom Chapter 2 Rev 10-27-87 Typeset 10-27-87 3 Chapter 2 ZMODEM Protocol 4 degrade performance. Some networks cannot transmit 1024 byte packets without flow control, which is difficult to apply without impairing the perfect transparency required by XMODEM. XMODEM-k adds garbage to received files. o YMODEM sends the file name, file length, and creation date at the beginning of each file, and allows optional 1024 byte blocks for improved throughput. The handling of files that are not a multiple of 1024 or 128 bytes is awkward, especially if the file length is not known in advance, or changes during transmission. The large number of non conforming and substandard programs claiming to support YMODEM further complicates its use. o YMODEM-g provides efficient batch file transfers, preserving exact file length and file modification date. YMODEM-g is a modification to YMODEM wherein ACKs for data blocks are not used. YMODEM-g is essentially insensitive to network delays. Because it does not support error recovery, YMODEM-g must be used hard wired or with a reliable link level protocol. Successful application at high speed requires cafeful attention to transparent flow control. When YMODEM-g detects a CRC error, data transfers are aborted. YMODEM-g is easy to implement because it closely resembles standard YMODEM. o WXMODEM, SEAlink, and MEGAlink have applied a subset of ZMODEM's techniques to "Classic XMODEM" to improve upon their suppliers' previous offerings. They provide good performance under ideal conditions. Another XMODEM "extension" is protocol cheating, such as Omen Technology's OverThruster(TM) and OverThruster II(TM). These improve XMODEM throughput under some conditions by compromising error recovery. The ZMODEM Protocol corrects the weaknesses described above while maintaining as much of XMODEM/CRC's simplicity and prior art as possible. 3. ZMODEM Protocol Design Criteria The design of a file transfer protocol is an engineering compromise between conflicting requirements: 3.1 Ease of Use o ZMODEM allows either program to initiate file transfers, passing commands and/or modifiers to the other program. o File names need be entered only once. o Menu selections are supported. Chapter 3 Rev 10-27-87 Typeset 10-27-87 4 Chapter 3 ZMODEM Protocol 5 o Wild Card names may be used with batch transfers. o Minimum keystrokes required to initiate transfers. o ZRQINIT frame sent by sending program can trigger automatic downloads. o ZMODEM can step down to YMODEM if the other end does not support ZMODEM.[1] 3.2 Throughput All file transfer protocols make tradeoffs between throughput, reliability, universality, and complexity according to the technology and knowledge base available to their designers. In the design of ZMODEM, three applications deserve special attention. o Network applications with significant delays (relative to character transmission time) and low error rate o Timesharing and buffered modem applications with significant delays and throughput that is quickly degraded by reverse channel traffic. ZMODEM's economy of reverse channel bandwidth allows modems that dynamically partition bandwidth between the two directions to operate at optimal speeds. Special ZMODEM features allow simple, efficient implementation on a wide variety of timesharing hosts. o Direct modem to modem communications with high error rate Unlike Sliding Windows Kermit, ZMODEM is not optimized for optimum throughput when error rate and delays are both high. This tradeoff markedly reduces code complexity and memory requirements. ZMODEM generally provides faster error recovery than network compatible XMODEM implementations. In the absence of network delays, rapid error recovery is possible, much faster than MEGAlink and network compatible versions of YMODEM and XMODEM. File transfers begin immediately regardless of which program is started first, without the 10 second delay associated with XMODEM. __________ 1. Provided the transmission medium accommodates X/YMODEM. Chapter 3 Rev 10-27-87 Typeset 10-27-87 5 Chapter 3 ZMODEM Protocol 6 3.3 Integrity and Robustness Once a ZMODEM session is begun, all transactions are protected with 16 or 32 bit CRC.[2] Complex proprietary techniques such as Cybernetic Data Recovery(TM)[3] are not needed for reliable transfers. An optional 32-bit CRC used as the frame check sequence in ADCCP (ANSI X3.66, also known as FIPS PUB 71 and FED-STD-1003, the U.S. versions of CCITT's X.25) is used when available. The 32 bit CRC reduces undetected errors by at least five orders of magnitude when properly applied (-1 preset, inversion). A security challenge mechanism guards against "Trojan Horse" messages written to mimic legitimate command or file downloads. 3.4 Ease of Implementation ZMODEM accommodates a wide variety of systems: o Microcomputers that cannot overlap disk and serial i/o o Microcomputers that cannot overlap serial send and receive o Computers and/or networks requiring XON/XOFF flow control o Computers that cannot check the serial input queue for the presence of data without having to wait for the data to arrive. Although ZMODEM provides "hooks" for multiple "threads", ZMODEM is not intended to replace link level protocols such as X.25. ZMODEM accommodates network and timesharing system delays by continuously transmitting data unless the receiver interrupts the sender to request retransmission of garbled data. ZMODEM in effect uses the entire file as a window.[4] Using the entire file as a window simplifies buffer management, avoiding the window overrun failure modes that affect MEGAlink, SuperKermit, and others. ZMODEM provides a general purpose application to application file transfer protocol which may be used directly or with with reliable link level __________ 2. Except for the CAN-CAN-CAN-CAN-CAN abort sequence which requires five successive CAN characters. 3. Unique to Professional-YAM and PowerCom 4. Streaming strategies are discussed in coming chapters. Chapter 3 Rev 10-27-87 Typeset 10-27-87 6 Chapter 3 ZMODEM Protocol 7 protocols such as X.25, MNP, Fastlink, etc. When used with X.25, MNP, Fastlink, etc., ZMODEM detects and corrects errors in the interfaces between error controlled media and the remainder of the communications link. ZMODEM was developed for the public domain under a Telenet contract. The ZMODEM protocol descriptions and the Unix rz/sz program source code are public domain. No licensing, trademark, or copyright restrictions apply to the use of the protocol, the Unix rz/sz source code and the ZMODEM name. 4. EVOLUTION OF ZMODEM In early 1986, Telenet funded a project to develop an improved public domain application to application file transfer protocol. This protocol would alleviate the throughput problems network customers were experiencing with XMODEM and Kermit file transfers. In the beginning, we thought a few modifications to XMODEM would allow high performance over packet switched networks while preserving XMODEM's simplicity. The initial concept would add a block number to the ACK and NAK characters used by XMODEM. The resultant protocol would allow the sender to send more than one block before waiting for a response. But how to add the block number to XMODEM's ACK and NAK? WXMODEM, SEAlink, MEGAlink and some other protocols add binary byte(s) to indicate the block number. Pure binary was unsuitable for ZMODEM because binary code combinations won't pass bidirectionally through some modems, networks and operating systems. Other operating systems may not be able to recognize something coming back[1] unless a break signal or a system dependent code or sequence is present. By the time all this and other problems with the simple ACK/NAK sequences mentioned above were corrected, XMODEM's simple ACK and NACK characters had evolved into a real packet. The Frog was riveting. Managing the window[2] was another problem. Experience gained in debugging The Source's SuperKermit protocol indicated a window size of about 1000 characters was needed at 1200 bps. High speed modems require a __________ 1. Without stopping for a response 2. The WINDOW is the data in transit between sender and receiver. Chapter 4 Rev 10-27-87 Typeset 10-27-87 7 Chapter 4 ZMODEM Protocol 8 window of 20000 or more characters for full throughput. Much of the SuperKermit's inefficiency, complexity and debugging time centered around its ring buffering and window management. There had to be a better way to get the job done. A sore point with XMODEM and its progeny is error recovery. More to the point, how can the receiver determine whether the sender has responded, or is ready to respond, to a retransmission request? XMODEM attacks the problem by throwing away characters until a certain period of silence. Too short a time allows a spurious pause in output (network or timesharing congestion) to masquerade as error recovery. Too long a timeout devastates throughput, and allows a noisy line to lock up the protocol. SuperKermit solves the problem with a distinct start of packet character (SOH). WXMODEM and ZMODEM use unique character sequences to delineate the start of frames. SEAlink and MEGAlink do not address this problem. A further error recovery problem arises in streaming protocols. How does the receiver know when (or if) the sender has recognized its error signal? Is the next packet the correct response to the error signal? Is it something left over "in the queue"? Or is this new subpacket one of many that will have to be discarded because the sender did not receive the error signal? How long should this continue before sending another error signal? How can the protocol prevent this from degenerating into an argument about mixed signals? SuperKermit uses selective retransmission, so it can accept any good packet it receives. Each time the SuperKermit receiver gets a data packet, it must decide which outstanding packet (if any) it "wants most" to receive, and asks for that one. In practice, complex software "hacks" are needed to attain acceptable robustness.[3] For ZMODEM, we decided to forgo the complexity of SuperKermit's packet assembly scheme and its associated buffer management logic and memory requirements. Another sore point with XMODEM and WXMODEM is the garbage added to files. This was acceptable with old CP/M files which had no exact length, but not with modern systems such as DOS and Unix. YMODEM uses file length information transmitted in the header block to trim the output file, but this causes data loss when transferring files that grow during a transfer. In some cases, the file length may be unknown, as when data is obtained from a process. Variable length data subpackets solve both of these __________ 3. For example, when SuperKermit encounters certain errors, the wndesr function is called to determine the next block to request. A burst of errors generates several wasteful requests to retransmit the same block. Chapter 4 Rev 10-27-87 Typeset 10-27-87 8 Chapter 4 ZMODEM Protocol 9 problems. Since some characters had to be escaped anyway, there wasn't any point wasting bytes to fill out a fixed packet length or to specify a variable packet length. In ZMODEM, the length of data subpackets is denoted by ending each subpacket with an escape sequence similar to BISYNC and HDLC. The end result is a ZMOEM header containing a "frame type", four bytes of supervisory information, and its own CRC. Data frames consist of a header followed by 1 or more data subpackets. In the absence of transmission errors, an entire file can be sent in one data frame. Since the sending system may be sensitive to numerous control characters or strip parity in the reverse data path, all of the headers sent by the receiver are sent in hex. A common lower level routine receives all headers, allowing the main program logic to deal with headers and data subpackets as objects. With equivalent binary (efficient) and hex (application friendly) frames, the sending program can send an "invitation to receive" sequence to activate the receiver without crashing the remote application with unexpected control characters. Going "back to scratch" in the protocol design presents an opportunity to steal good ideas from many sources and to add a few new ones. From Kermit and UUCP comes the concept of an initial dialog to exchange system parameters. ZMODEM generalizes Compuserve B Protocol's host controlled transfers to single command AutoDownload and command downloading. A Security Challenge discourages password hackers and Trojan Horse authors from abusing ZMODEM's power. We were also keen to the pain and $uffering of legions of telecommunicators whose file transfers have been ruined by communications and timesharing faults. ZMODEM's file transfer recovery and advanced file management are dedicated to these kindred comrades. After ZMODEM had been operational a short time, Earl Hall pointed out the obvious: ZMODEM's user friendly AutoDownload was almost useless if the user must assign transfer options to each of the sending and receiving programs. Now, transfer options may be specified to/by the sending program, which passes them to the receiving program in the ZFILE header. Chapter 5 Rev 10-27-87 Typeset 10-27-87 9 Chapter 5 ZMODEM Protocol 10 5. ROSETTA STONE Here are some definitions which reflect current vernacular in the computer media. The attempt here is identify the file transfer protocol rather than specific programs. FRAME A ZMODEM frame consists of a header and 0 or more data subpackets. XMODEM refers to the original 1977 file transfer etiquette introduced by Ward Christensen's MODEM2 program. It's also called the MODEM or MODEM2 protocol. Some who are unaware of MODEM7's unusual batch file mode call it MODEM7. Other aliases include "CP/M Users's Group" and "TERM II FTP 3". This protocol is supported by most communications programs because it is easy to implement. XMODEM/CRC replaces XMODEM's 1 byte checksum with a two byte Cyclical Redundancy Check (CRC-16), improving error detection. XMODEM-1k Refers to XMODEM-CRC with optional 1024 byte blocks. YMODEM refers to the XMODEM/CRC protocol with batch transmission and optional 1024 byte blocks as described in YMODEM.DOC.[1] 6. ZMODEM REQUIREMENTS ZMODEM requires an 8 bit transfer medium.[1] ZMODEM escapes network control characters to allow operation with packet switched networks. In general, ZMODEM operates over any path that supports XMODEM, and over many that don't. To support full streaming,[2] the transmission path should either assert flow control or pass full speed transmission without loss of data. Otherwise the ZMODEM sender must manage the window size. 6.1 File Contents 6.1.1 Binary Files ZMODEM places no constraints on the information content of binary files, except that the number of bits in the file must be a multiple of 8. __________ 1. Available on TeleGodzilla as part of YZMODEM.ZOO 1. The ZMODEM design allows encoded packets for less transparent media. 2. With XOFF and XON, or out of band flow control such as X.25 or CTS Chapter 6 Rev 10-27-87 Typeset 10-27-87 10 Chapter 6 ZMODEM Protocol 11 6.1.2 Text Files Since ZMODEM is used to transfer files between different types of computer systems, text files must meet minimum requirements if they are to be readable on a wide variety of systems and environments. Text lines consist of printing ASCII characters, spaces, tabs, and backspaces. 6.1.2.1 ASCII End of Line The ASCII code definition allows text lines terminated by a CR/LF (015, 012) sequence, or by a NL (012) character. Lines logically terminated by a lone CR (013) are not ASCII text. A CR (013) without a linefeed implies overprinting, and is not acceptable as a logical line separator. Overprinted lines should print all important characters in the last pass to allow CRT displays to display meaningful text. Overstruck characters may be generated by backspacing or by overprinting the line with CR (015) not followed by LF. Overstruck characters generated with backspaces should be sent with the most important character last to accommodate CRT displays that cannot overstrike. The sending program may use the ZCNL bit to force the receiving program to convert the received end of line to its local end of line convention.[3] __________ 3. Files that have been translated in such a way as to modify their length cannot be updated with the ZCRECOV Conversion Option. Chapter 6 Rev 10-27-87 Typeset 10-27-87 11 Chapter 6 ZMODEM Protocol 12 7. ZMODEM BASICS 7.1 Packetization ZMODEM frames differ somewhat from XMODEM blocks. XMODEM blocks are not used for the following reasons: o Block numbers are limited to 256 o No provision for variable length blocks o Line hits corrupt protocol signals, causing failed file transfers. In particular, modem errors sometimes generate false block numbers, false EOTs and false ACKs. False ACKs are the most troublesome as they cause the sender to lose synchronization with the receiver. State of the art programs such as Professional-YAM and ZCOMM overcome some of these weaknesses with clever proprietary code, but a stronger protocol is desired. o It is difficult to determine the beginning and ends of XMODEM blocks when line hits cause a loss of synchronization. This precludes rapid error recovery. 7.2 Link Escape Encoding ZMODEM achieves data transparency by extending the 8 bit character set (256 codes) with escape sequences based on the ZMODEM data link escape character ZDLE.[1] Link Escape coding permits variable length data subpackets without the overhead of a separate byte count. It allows the beginning of frames to be detected without special timing techniques, facilitating rapid error recovery. Link Escape coding does add some overhead. The worst case, a file consisting entirely of escaped characters, would incur a 50% overhead. The ZDLE character is special. ZDLE represents a control sequence of some sort. If a ZDLE character appears in binary data, it is prefixed with ZDLE, then sent as ZDLEE. The value for ZDLE is octal 030 (ASCII CAN). This particular value was chosen to allow a string of 5 consecutive CAN characters to abort a ZMODEM __________ 1. This and other constants are defined in the zmodem.h include file. Please note that constants with a leading 0 are octal constants in C. Chapter 7 Rev 10-27-87 Typeset 10-27-87 12 Chapter 7 ZMODEM Protocol 13 session, compatible with YMODEM session abort. Since CAN is not used in normal terminal operations, interactive applications and communications programs can monitor the data flow for ZDLE. The following characters can be scanned to detect the ZRQINIT header, the invitation to automatically download commands or files. Receipt of five successive CAN characters will abort a ZMODEM session. Eight CAN characters are sent. The receiving program decodes any sequence of ZDLE followed by a byte with bit 6 set and bit 5 reset (upper case letter, either parity) to the equivalent control character by inverting bit 6. This allows the transmitter to escape any control character that cannot be sent by the communications medium. In addition, the receiver recognizes escapes for 0177 and 0377 should these characters need to be escaped. ZMODEM software escapes ZDLE, 020, 0220, 021, 0221, 023, and 0223. If preceded by 0100 or 0300 (@), 015 and 0215 are also escaped to protect the Telenet command escape CR-@-CR. The receiver ignores 021, 0221, 023, and 0223 characters in the data stream. The ZMODEM routines in zm.c accept an option to escape all control characters, to allow operation with less transparent networks. This option can be given to either the sending or receiving program. 7.3 Header All ZMODEM frames begin with a header which may be sent in binary or HEX form. ZMODEM uses a single routine to recognize binary and hex headers. Either form of the header contains the same raw information: o A type byte[2] [3] o Four bytes of data indicating flags and/or numeric quantities depending on the frame type __________ 2. The frame types are cardinal numbers beginning with 0 to minimize state transition table memory requirements. 3. Future extensions to ZMODEM may use the high order bits of the type byte to indicate thread selection. Chapter 7 Rev 10-27-87 Typeset 10-27-87 13 Chapter 7 ZMODEM Protocol 14 Figure 1. Order of Bytes in Header TYPE: frame type F0: Flags least significant byte P0: file Position least significant P3: file Position most significant TYPE F3 F2 F1 F0 ------------------- TYPE P0 P1 P2 P3 7.3.1 16 Bit CRC Binary Header A binary header is sent by the sending program to the receiving program. ZDLE encoding accommodates XON/XOFF flow control. A binary header begins with the sequence ZPAD, ZDLE, ZBIN. The frame type byte is ZDLE encoded. The four position/flags bytes are ZDLE encoded. A two byte CRC of the frame type and position/flag bytes is ZDLE encoded. 0 or more binary data subpackets with 16 bit CRC will follow depending on the frame type. The function zsbhdr transmits a binary header. The function zgethdr receives a binary or hex header. Figure 2. 16 Bit CRC Binary Header * ZDLE A TYPE F3/P0 F2/P1 F1/P2 F0/P3 CRC-1 CRC-2 7.3.2 32 Bit CRC Binary Header A "32 bit CRC" Binary header is similar to a Binary Header, except the ZBIN (A) character is replaced by a ZBIN32 (C) character, and four characters of CRC are sent. 0 or more binary data subpackets with 32 bit CRC will follow depending on the frame type. The common variable Txfcs32 may be set TRUE for 32 bit CRC iff the receiver indicates the capability with the CANFC32 bit. The zgethdr, zsdata and zrdata functions automatically adjust to the type of Frame Check Sequence being used. Figure 3. 32 Bit CRC Binary Header * ZDLE C TYPE F3/P0 F2/P1 F1/P2 F0/P3 CRC-1 CRC-2 CRC-3 CRC-4 7.3.3 HEX Header The receiver sends responses in hex headers. The sender also uses hex headers when they are not followed by binary data subpackets. Chapter 7 Rev 10-27-87 Typeset 10-27-87 14 Chapter 7 ZMODEM Protocol 15 Hex encoding protects the reverse channel from random control characters. The hex header receiving routine ignores parity. Use of Kermit style encoding for control and paritied characters was considered and rejected because of increased possibility of interacting with some timesharing systems' line edit functions. Use of HEX headers from the receiving program allows control characters to be used to interrupt the sender when errors are detected. A HEX header may be used in place of a binary header wherever convenient. If a data packet follows a HEX header, it is protected with CRC-16. A hex header begins with the sequence ZPAD, ZPAD, ZDLE, ZHEX. The zgethdr routine synchronizes with the ZPAD-ZDLE sequence. The extra ZPAD character allows the sending program to detect an asynchronous header (indicating an error condition) and then call zgethdr to receive the header. The type byte, the four position/flag bytes, and the 16 bit CRC thereof are sent in hex using the character set 01234567890abcdef. Upper case hex digits are not allowed; they false trigger XMODEM and YMODEM programs. Since this form of hex encoding detects many patterns of errors, especially missing characters, a hex header with 32 bit CRC has not been defined. A carriage return and line feed are sent with HEX headers. The receive routine expects to see at least one of these characters, two if the first is CR. The CR/LF aids debugging from printouts, and helps overcome certain operating system related problems. An XON character is appended to all HEX packets except ZACK and ZFIN. The XON releases the sender from spurious XOFF flow control characters generated by line noise, a common occurrence. XON is not sent after ZACK headers to protect flow control in streaming situations. XON is not sent after a ZFIN header to allow clean session cleanup. 0 or more data subpackets will follow depending on the frame type. The function zshhdr sends a hex header. Figure 4. HEX Header * * ZDLE B TYPE F3/P0 F2/P1 F1/P2 F0/P3 CRC-1 CRC-2 CR LF XON (TYPE, F3...F0, CRC-1, and CRC-2 are each sent as two hex digits.) Chapter 7 Rev 10-27-87 Typeset 10-27-87 15 Chapter 7 ZMODEM Protocol 16 7.4 Binary Data Subpackets Binary data subpackets immediately follow the associated binary header packet. A binary data packet contains 0 to 1024 bytes of data. Recommended length values are 256 bytes below 2400 bps, 512 at 2400 bps, and 1024 above 4800 bps or when the data link is known to be relatively error free.[4] No padding is used with binary data subpackets. The data bytes are ZDLE encoded and transmitted. A ZDLE and frameend are then sent, followed by two or four ZDLE encoded CRC bytes. The CRC accumulates the data bytes and frameend. The function zsdata sends a data subpacket. The function zrdata receives a data subpacket. 7.5 ASCII Encoded Data Subpacket The format of ASCII Encoded data subpackets is not currently specified. These could be used for server commands, or main transfers in 7 bit environments. 8. PROTOCOL TRANSACTION OVERVIEW As with the XMODEM recommendation, ZMODEM timing is receiver driven. The transmitter should not time out at all, except to abort the program if no headers are received for an extended period of time, say one minute.[1] 8.1 Session Startup To start a ZMODEM file transfer session, the sending program is called with the names of the desired file(s) and option(s). The sending program may send the string "rz\r" to invoke the receiving program from a possible command mode. The "rz" followed by carriage return activates a ZMODEM receive program or command if it were not already active. The sender may then display a message intended for human consumption, such __________ 4. Strategies for adjusting the subpacket length for optimal results based on real time error rates are still evolving. Shorter subpackets speed error detection but increase protocol overhead slightly. 1. Special considerations apply when sending commands. Chapter 8 Rev 10-27-87 Typeset 10-27-87 16 Chapter 8 ZMODEM Protocol 17 as a list of the files requested, etc. Then the sender may send a ZRQINIT header. The ZRQINIT header causes a previously started receive program to send its ZRINIT header without delay. In an interactive or conversational mode, the receiving application may monitor the data stream for ZDLE. The following characters may be scanned for B00 indicating a ZRQINIT header, a command to download a command or data. The sending program awaits a command from the receiving program to start file transfers. If a "C", "G", or NAK is received, an XMODEM or YMODEM file transfer is indicated, and file transfer(s) use the YMODEM protocol. Note: With ZMODEM and YMODEM, the sending program provides the file name, but not with XMODEM. In case of garbled data, the sending program can repeat the invitation to receive a number of times until a session starts. When the ZMODEM receive program starts, it immediately sends a ZRINIT header to initiate ZMODEM file transfers, or a ZCHALLENGE header to verify the sending program. The receive program resends its header at response time (default 10 second) intervals for a suitable period of time (40 seconds total) before falling back to YMODEM protocol. If the receiving program receives a ZRQINIT header, it resends the ZRINIT header. If the sending program receives the ZCHALLENGE header, it places the data in ZP0...ZP3 in an answering ZACK header. If the receiving program receives a ZRINIT header, it is an echo indicating that the sending program is not operational. Eventually the sending program correctly receives the ZRINIT header. The sender may then send an optional ZSINIT frame to define the receiving program's Attn sequence, or to specify complete control character escaping.[2] If the ZSINIT header specifies ESCCTL or ESC8, a HEX header is used, and the receiver activates the specified ESC modes before reading the following data subpacket. The receiver sends a ZACK header in response, optionally containing the __________ 2. If the receiver specifies the same or higher level of escaping, the ZSINIT frame need not be sent unless an Attn sequence is needed. Chapter 8 Rev 10-27-87 Typeset 10-27-87 17 Chapter 8 ZMODEM Protocol 18 serial number of the receiving program, or 0. 8.2 File Transmission The sender then sends a ZFILE header with ZMODEM Conversion, Management, and Transport options[3] followed by a ZCRCW data subpacket containing the file name, file length, modification date, and other information identical to that used by YMODEM Batch. The receiver examines the file name, length, and date information provided by the sender in the context of the specified transfer options, the current state of its file system(s), and local security requirements. The receiving program should insure the pathname and options are compatible with its operating environment and local security requirements. The receiver may respond with a ZSKIP header, which makes the sender proceed to the next file (if any) in the batch. If the receiver has a file with the same name and length, it may respond with a ZCRC header, which requires the sender to perform a 32 bit CRC on the file and transmit the complement of the CRC in a ZCRC header.[4] The receiver uses this information to determine whether to accept the file or skip it. This sequence is triggered by the ZMCRC Management Option. A ZRPOS header from the receiver initiates transmission of the file data starting at the offset in the file specified in the ZRPOS header. Normally the receiver specifies the data transfer to begin begin at offset 0 in the file. The receiver may start the transfer further down in the file. This allows a file transfer interrupted by a loss or carrier or system crash to be completed on the next connection without requiring the entire file to be retransmitted.[5] If downloading a file from a timesharing system that becomes sluggish, the transfer can be interrupted and resumed later with no loss of data. The sender sends a ZDATA binary header (with file position) followed by one or more data subpackets. __________ 3. See below, under ZFILE header type. 4. The crc is initialized to 0xFFFFFFFF. 5. This does not apply to files that have been translated. Chapter 8 Rev 10-27-87 Typeset 10-27-87 18 Chapter 8 ZMODEM Protocol 19 The receiver compares the file position in the ZDATA header with the number of characters successfully received to the file. If they do not agree, a ZRPOS error response is generated to force the sender to the right position within the file.[6] A data subpacket terminated by ZCRCG and CRC does not elicit a response unless an error is detected; more data subpacket(s) follow immediately. ZCRCQ data subpackets expect a ZACK response with the receiver's file offset if no error, otherwise a ZRPOS response with the last good file offset. Another data subpacket continues immediately. ZCRCQ subpackets are not used if the receiver does not indicate FDX ability with the CANFDX bit. ZCRCW data subpackets expect a response before the next frame is sent. If the receiver does not indicate overlapped I/O capability with the CANOVIO bit, or sets a buffer size, the sender uses the ZCRCW to allow the receiver to write its buffer before sending more data. A zero length data frame may be used as an idle subpacket to prevent the receiver from timing out in case data is not immediately available to the sender. In the absence of fatal error, the sender eventually encounters end of file. If the end of file is encountered within a frame, the frame is closed with a ZCRCE data subpacket which does not elicit a response except in case of error. The sender sends a ZEOF header with the file ending offset equal to the number of characters in the file. The receiver compares this number with the number of characters received. If the receiver has received all of the file, it closes the file. If the file close was satisfactory, the receiver responds with ZRINIT. If the receiver has not received all the bytes of the file, the receiver ignores the ZEOF because a new ZDATA is coming. If the receiver cannot properly close the file, a ZFERR header is sent. After all files are processed, any further protocol errors should not prevent the sending program from returning with a success status. __________ 6. If the ZMSPARS option is used, the receiver instead seeks to the position given in the ZDATA header. Chapter 8 Rev 10-27-87 Typeset 10-27-87 19 Chapter 8 ZMODEM Protocol 20 8.3 Session Cleanup The sender closes the session with a ZFIN header. The receiver acknowledges this with its own ZFIN header. When the sender receives the acknowledging header, it sends two characters, "OO" (Over and Out) and exits to the operating system or application that invoked it. The receiver waits briefly for the "O" characters, then exits whether they were received or not. 8.4 Session Abort Sequence If the receiver is receiving data in streaming mode, the Attn sequence is executed to interrupt data transmission before the Cancel sequence is sent. The Cancel sequence consists of eight CAN characters and ten backspace characters. ZMODEM only requires five Cancel characters, the other three are "insurance". The trailing backspace characters attempt to erase the effects of the CAN characters if they are received by a command interpreter. static char canistr[] = { 24,24,24,24,24,24,24,24,8,8,8,8,8,8,8,8,8,8,0 }; Chapter 8 Rev 10-27-87 Typeset 10-27-87 20 Chapter 8 ZMODEM Protocol 21 9. STREAMING TECHNIQUES / ERROR RECOVERY It is a fact of life that no single method of streaming is applicable to a majority of today's computing and telecommunications environments. ZMODEM provides several data streaming methods selected according to the limitations of the sending environment, receiving environment, and transmission channel(s). 9.1 Full Streaming with Sampling If the receiver can overlap serial I/O with disk I/O, and if the sender can sample the reverse channel for the presence of data without having to wait, full streaming can be used with no Attn sequence required. The sender begins data transmission with a ZDATA header and continuous ZCRCG data subpackets. When the receiver detects an error, it executes the Attn sequence and then sends a ZRPOS header with the correct position within the file. At the end of each transmitted data subpacket, the sender checks for the presence of an error header from the receiver. To do this, the sender samples the reverse data stream for the presence of either a ZPAD or CAN character.[1] Flow control characters (if present) are acted upon. Other characters (indicating line noise) increment a counter which is reset whenever the sender waits for a header from the receiver. If the counter overflows, the sender sends the next data subpacket as ZCRCW, and waits for a response. ZPAD indicates some sort of error header from the receiver. A CAN suggests the user is attempting to "stop the bubble machine" by keyboarding CAN characters. If one of these characters is seen, an empty ZCRCE data subpacket is sent. Normally, the receiver will have sent an ZRPOS or other error header, which will force the sender to resume transmission at a different address, or take other action. In the unlikely event the ZPAD or CAN character was spurious, the receiver will time out and send a ZRPOS header.[2] Then the receiver's response header is read and acted upon.[3] __________ 1. The call to rdchk() in sz.c performs this function. 2. The obvious choice of ZCRCW packet, which would trigger an ZACK from the receiver, is not used because multiple in transit frames could result if the channel has a long propagation delay. 3. The call to getinsync() in sz.c performs this function. Chapter 9 Rev 10-27-87 Typeset 10-27-87 21 Chapter 9 ZMODEM Protocol 22 A ZRPOS header resets the sender's file offset to the correct position. If possible, the sender should purge its output buffers and/or networks of all unprocessed output data, to minimize the amount of unwanted data the receiver must discard before receiving data starting at the correct file offset. The next transmitted data frame should be a ZCRCW frame followed by a wait to guarantee complete flushing of the network's memory. If the receiver gets a ZACK header with an address that disagrees with the sender address, it is ignored, and the sender waits for another header. A ZFIN, ZABORT, or TIMEOUT terminates the session; a ZSKIP terminates the processing of this file. The reverse channel is then sampled for the presence of another header from the receiver.[4] if one is detected, the getinsync() function is again called to read another error header. Otherwise, transmission resumes at the (possibly reset) file offset with a ZDATA header followed by data subpackets. 9.1.1 Window Management When sending data through a network, some nodes of the network store data while it is transferred to the receiver. 7000 bytes and more of transient storage have been observed. Such a large amount of storage causes the transmitter to "get ahead" of the reciever. This can be fatal with MEGAlink and other protocols that depend on timely notification of errors from the receiver. This condition is not fatal with ZMODEM, but it does slow error recovery. To manage the window size, the sending program uses ZCRCQ data subpackets to trigger ZACK headers from the receiver. The returning ZACK headers inform the sender of the receiver's progress. When the window size (current transmitter file offset - last reported receiver file offset) exceeds a specified value, the sender waits for a ZACK[5] packet with a receiver file offset that reduces the window size. Unix sz versions beginning with May 9 1987 control the window size with the "-w N" option, where N is the maximum window size. Pro-YAM, ZCOMM and DSZ versions beginning with May 9 1987 control the window size with "zmodem pwN". This is compatible with previous versions of these programs.[6] __________ 4. If sampling is possible. 5. ZRPOS and other error packets are handled normally. 6. When used with modems or networks that simultaneously assert flow Chapter 9 Rev 10-27-87 Typeset 10-27-87 22 Chapter 9 ZMODEM Protocol 23 9.2 Full Streaming with Reverse Interrupt The above method cannot be used if the reverse data stream cannot be sampled without entering an I/O wait. An alternate method is to instruct the receiver to interrupt the sending program when an error is detected. The receiver can interrupt the sender with a control character, break signal, or combination thereof, as specified in the Attn sequence. After executing the Attn sequence, the receiver sends a hex ZRPOS header to force the sender to resend the lost data. When the sending program responds to this interrupt, it reads a HEX header (normally ZRPOS) from the receiver and takes the action described in the previous section. The Unix sz.c program uses a setjmp/longjmp call to catch the interrupt generated by the Attn sequence. Catching the interrupt activates the getinsync() function to read the receiver's error header and take appropriate action. When compiled for standard SYSTEM III/V Unix, sz.c uses an Attn sequence of Ctrl-C followed by a 1 second pause to interrupt the sender, then give the sender (Unix) time to prepare for the receiver's error header. 9.3 Full Streaming with Sliding Window If none of the above methods is applicable, hope is not yet lost. If the sender can buffer responses from the receiver, the sender can use ZCRCQ data subpackets to get ACKs from the receiver without interrupting the transmission of data. After a sufficient number of ZCRCQ data subpackets have been sent, the sender can read one of the headers that should have arrived in its receive interrupt buffer. A problem with this method is the possibility of wasting an excessive amount of time responding to the receiver's error header. It may be possible to program the receiver's Attn sequence to flush the sender's interrupt buffer before sending the ZRPOS header. __________________________________________________________________________ control with XON and XOFF characters and pass XON characters that violate flow control, the receiving program should have a revision date of May 9 or later. Chapter 9 Rev 10-27-87 Typeset 10-27-87 23 Chapter 9 ZMODEM Protocol 24 9.4 Full Streaming over Error Free Channels File transfer protocols predicated on the existence of an error free end to end communications channel have been proposed from time to time. Such channels have proven to be more readily available in theory than in actuality. The frequency of undetected errors increases when modem scramblers have more bits than the error detecting CRC. A ZMODEM sender assuming an error free channel with end to end flow control can send the entire file in one frame without any checking of the reverse stream. If this channel is completely transparent, only ZDLE need be escaped. The resulting protocol overhead for average long files is less than one per cent.[7] 9.5 Segmented Streaming If the receiver cannot overlap serial and disk I/O, it uses the ZRINIT frame to specify a buffer length which the sender will not overflow. The sending program sends a ZCRCW data subpacket and waits for a ZACK header before sending the next segment of the file. If the sending program supports reverse data stream sampling or interrupt, error recovery will be faster (on average) than a protocol (such as YMODEM) that sends large blocks. A sufficiently large receiving buffer allows throughput to closely approach that of full streaming. For example, 16kb segmented streaming adds about 3 per cent to full streaming ZMODEM file transfer times when the round trip delay is five seconds. 10. ATTENTION SEQUENCE The receiving program sends the Attn sequence whenever it detects an error and needs to interrupt the sending program. The default Attn string value is empty (no Attn sequence). The receiving program resets Attn to the empty default before each transfer session. The sender specifies the Attn sequence in its optional ZSINIT frame. The Attn string is terminated with a null. __________ 7. One in 256 for escaping ZDLE, about two (four if 32 bit CRC is used) in 1024 for data subpacket CRC's Chapter 10 Rev 10-27-87 Typeset 10-27-87 24 Chapter 10 ZMODEM Protocol 25 Two meta-characters perform special functions: o \335 (octal) Send a break signal o \336 (octal) Pause one second 11. FRAME TYPES The numeric values for the values shown in boldface are given in zmodem.h. Unused bits and unused bytes in the header (ZP0...ZP3) are set to 0. 11.1 ZRQINIT Sent by the sending program, to trigger the receiving program to send its ZRINIT header. This avoids the aggravating startup delay associated with XMODEM and Kermit transfers. The sending program may repeat the receive invitation (including ZRQINIT) if a response is not obtained at first. ZF0 contains ZCOMMAND if the program is attempting to send a command, 0 otherwise. 11.2 ZRINIT Sent by the receiving program. ZF0 and ZF1 contain the bitwise or of the receiver capability flags: #define CANCRY 8 /* Receiver can decrypt */ #define CANFDX 01 /* Rx can send and receive true FDX */ #define CANOVIO 02 /* Rx can receive data during disk I/O */ #define CANBRK 04 /* Rx can send a break signal */ #define CANCRY 010 /* Receiver can decrypt */ #define CANLZW 020 /* Receiver can uncompress */ #define CANFC32 040 /* Receiver can use 32 bit Frame Check */ #define ESCCTL 0100 /* Receiver expects ctl chars to be escaped