Chaos RTAPE protocol
The Chaos RTAPE protocol is used to provide a remote tape server on a Chaos network. The protocol is structured as a sequence of messages called a record stream, and each message specifies an operation to be performed or data to transmit.
A description of the protocol—as reverse engineered from ITS' DUMP backup program and the MIT/Symbolics RTAPE.c server for BSD Unix—follows.
Record stream
Once a connection has been established, the server sends a packet with the string RECORD STREAM VERSION 1. The client responds with the same string. Presumably the parties can negotiate another version, but so far only version 1 is known.
After that, both sides will exchange messages. Each message begins with an octet which is the opcode. After that comes two octets, in big-endian order, which is the size of the data that follows. Finally come as many octets of data as specified by the size.
The data portion may be an ASCII text string, or binary data to transmit tape blocks. If text data has several fields, they are separated by a space characters; numbers are in decimal with an optional minus sign.
Operations
Client-to-server opcodes:
| Opcode | Name | Data |
|---|---|---|
| 1 | Login | ? |
| 2 | Mount | Text: type reel drive size density options |
| 3 | Probe | Binary: two octets with an ID |
| 4 | Read | Nothing (read all records until file mark), or text: number-of-records |
| 5 | Write | Text: block-size (decimal) |
| 6 | Rewind | Nothing |
| 7 | Rewind-sync | Nothing |
| 8 | Rewind-unload | Nothing |
| 9 | Space file | Text: number-of-files |
| 10 | Space record | Text: number-of-records |
| 12 | Write file mark | Nothing |
| 13 | Close | Nothing |
Server-to-client opcodes:
| Opcode | Name | Data |
|---|---|---|
| 33 | Login response | ? |
| 34 | Read data | Binary: tape block data |
| 35 | Read file mark | Nothing |
| 36 | Status | See below |
Status messages may be solicited by the Probe opcode, or the server can send them at any time to signal an error.
The Read operation may transmit several tape blocks, but the client can interrupt this at any time by sending a Probe.
The data in a mount message needs further explanation:
| Field | Description |
|---|---|
| Type | Mode: READ, WRITE, or BOTH
|
| Reel | ? |
| Drive | The name of the tape drive, as understood by the server |
| Size | (Preferred? max?) tape block size |
| Density | Tape density, in bytes per inch |
| Options | A space-separated list. May include NOREWIND and OFFLINE
|
Status information
The Status message has this data:
| Offset | Size | Description | Data |
|---|---|---|---|
| 0 | 1 | Version | Only 1 is known
|
| 1 | 2 | ID | Matching the Probe ID (Solicited flag set),or 0
|
| 3 | 3 | Number of blocks read | |
| 6 | 3 | Number of blocks skipped | |
| 9 | 3 | Number of blocks discarded | |
| 12 | 1 | Last operation received | |
| 13 | 2 | Density | |
| 15 | 2 | Retries in last operation | |
| 17 | 1 | Length of drive name | |
| 18 | 16 | Name of drive | |
| 34 | 2 | Flags | |
| 36 | ? | Optional error message |
The two-octet flags field, when taken as a little-endian number:
| Bit | Description |
|---|---|
| 0 | Solicited, e.g. requested by Probe
|
| 1 | Drive is at beginning of tape |
| 2 | Drive is past end of tape |
| 3 | Last operation reached a file mark |
| 4 | Not logged in |
| 5 | Tape is mounted |
| 6 | Error message follows |
| 7 | Hard error |
| 8 | Soft error |
| 9 | Drive is offline |
