RFC 1184
Telnet Linemode Option
Telnet Linemode Option ๐
okay bestie, let's get the basics out of the way first ๐ this is RFC 1184, called Telnet Linemode Option. It was written by D. Borman (Editor) at Cray Research, Inc., for the Network Working Group and the Telnet Working Group, and it's dated October 1990. Its status is Draft Standard, an IAB standards track protocol, and it obsoletes RFC 1116 ๐ฆ
Status of This Memo โจ
this memo is a Draft Standard for the Internet community, and it's basically begging everyone to discuss it and suggest improvements ๐ this RFC specifies an IAB standards track protocol for the Internet community. for the current standardization state and status of this protocol, check the current edition of the "IAB Official Protocol Standards." distribution of this memo is unlimited ๐ธ
Changes from RFC1116 ๐
ngl this part is so good ๐ญ two new mode bits joined the party, SOFT_TAB and LIT_ECHO. they let the server give the client some advice on how to echo tabs and non-printable characters ๐ง
there are also a bunch of new special character mappings for cursor motion when visual editing is supported:
- move cursor one character left/right:
SLC_MCL/SLC_MCR - move cursor one word left/right:
SLC_MCWL/SLC_MCWR - move cursor to beginning/end of line:
SLC_MCBOL/SLC_MCEOL - enter insert/overstrike mode:
SLC_INSRT/SLC_OVER - erase one character/word to the right:
SLC_ECR/SLC_EWR - erase to the beginning/end of the line:
SLC_EBOL/SLC_EEOL
Overview โจ
okay so Linemode Telnet is a way of doing terminal character processing on the client side of a Telnet connection. while you're in Linemode with editing turned on for the local side, network traffic drops to a couple of packets per command line, instead of a couple of packets per character typed. this is SO useful on long delay networks, because the user gets local response time while typing the command line, and only feels the network delay after the command is sent ๐ฅ it also saves money on networks that charge per packet ๐ please send comments to the telnet-ietf@cray.com mailing list ๐ซถ
Table of Contents ๐
1. Command Names and Codes 2
2. Command Meanings 3
2.1 The LINEMODE function 3
2.2 LINEMODE suboption MODE 4
2.3 LINEMODE suboption FORWARDMASK 5
2.4 LINEMODE suboption SLC, Set Local Characters 6
2.5 New control characters 10
3. Default Specification 11
4. Motivation 11
5. Implementation Rules 13
5.1 User Interface 13
5.2 End of line terminators 14
5.3 Output processing 14
5.4 A terminal driver in Telnet? 14
5.5 Setting of Local Characters 14
5.6 FORWARDMASK and SLC_FORW1 and SLC_FORW2 15
5.7 Valid and invalid modes and values 16
5.8 Flushing input and output 16
5.9 State diagram for SLC 18
5.10 Example of a connection 19
6. Other Telnet options and RFCs 22
7. Security Considerations 23
8. Author's Address 23
1. Command Names and Codes ๐
here's the big list of codes, all in one spot so you can find them fast ๐
LINEMODE 34
MODE 1
EDIT 1
TRAPSIG 2
MODE_ACK 4
SOFT_TAB 8
LIT_ECHO 16
FORWARDMASK 2
SLC 3
SLC_SYNCH 1
SLC_BRK 2
SLC_IP 3
SLC_AO 4
SLC_AYT 5
SLC_EOR 6
SLC_ABORT 7
SLC_EOF 8
SLC_SUSP 9
SLC_EC 10
SLC_EL 11
SLC_EW 12
SLC_RP 13
SLC_LNEXT 14
SLC_XON 15
SLC_XOFF 16
SLC_FORW1 17
SLC_FORW2 18
SLC_MCL 19
SLC_MCR 20
SLC_MCWL 21
SLC_MCWR 22
SLC_MCBOL 23
SLC_MCEOL 24
SLC_INSRT 25
SLC_OVER 26
SLC_ECR 27
SLC_EWR 28
SLC_EBOL 29
SLC_EEOL 30
SLC_DEFAULT 3
SLC_VALUE 2
SLC_CANTCHANGE 1
SLC_NOSUPPORT 0
SLC_LEVELBITS 3
SLC_ACK 128
SLC_FLUSHIN 64
SLC_FLUSHOUT 32
EOF 236
SUSP 237
ABORT 238
2. Command Meanings ๐ฆ
2.1 The LINEMODE function โจ
IAC WILL LINEMODE
the sender of this command requests permission to begin sub-negotiation of the editing/signaling status. this should only be sent by the client side of the connection ๐ง
IAC WONT LINEMODE
the sender of this command demands that sub-negotiation of the editing/signaling status not be allowed ๐
IAC DO LINEMODE
the sender of this command requests that the remote side begin sub-negotiation of the editing/signaling status. this should only be sent by the server side of the connection ๐
IAC DONT LINEMODE
the sender of this command demands that the remote side not begin sub-negotiation of the editing/signaling status ๐
2.2 LINEMODE suboption MODE ๐
this is the one that does all the heavy lifting, so pay attention ๐ฅ
IAC SB LINEMODE MODE mask IAC SE
the sender of this command confirms, or requests permission for, a switch to the mode defined by "mask" โจ
the "mask" is a bit mask of the various modes the connection can be in. under normal operation, the server side of the connection initiates mode changes, and the client confirms them. the currently defined modes are:
EDIT: when set, the client side of the connection should process all input lines, performing any editing functions, and only send completed lines to the remote side. when unset, the client side should not process any input from the user, and the server side should take care of all character processing that needs to be done ๐TRAPSIG: when set, the client side should translate appropriate interrupts/signals to their Telnet equivalent. (these would be IP, BRK, AYT, ABORT, EOF, and SUSP.) when unset, the client should pass interrupts/signals as their normal ASCII values ๐ฅFLOW: logically, this belongs in the "mask". however, it would overlap the Telnet TOGGLE-FLOW-CONTROL option, so the Telnet TOGGLE-FLOW-CONTROL option is used instead. when DO/WILL LINEMODE is negotiated, DO/WILL TOGGLE-FLOW-CONTROL should also be negotiated. see RFC 1080, "Telnet Remote Flow Control", for correct usage ๐ธECHO: logically, this belongs in the "mask" too. but it would overlap the Telnet ECHO option, so the Telnet ECHO option is used instead. the client side should never negotiate "WILL ECHO". when the server has negotiated "WILL ECHO", the client should not echo data typed by the user back to the user. when the server has negotiated "WONT ECHO", the client is responsible for echoing data typed by the user back to the user. see RFC 857, "Telnet ECHO OPTION", for a complete discussion on the use of the Telnet ECHO option ๐ฆSOFT_TAB: when set, the client side should expand the Horizontal Tab (HT) code, USASCII 9, into the appropriate number of spaces to move the printer to the next horizontal tab stop. when unset, the client side should allow the Horizontal Tab code to pass through un-modified ๐LIT_ECHO: when set, if the client side is echoing a non-printable character that the user has typed to the user's screen, the character should be echoed as the literal character. if theLIT_ECHObit is not set, the client side may echo the character in any manner it desires. (many systems echo unprintable characters as two-character sequences, for example, they'll echo "^A" for an ASCII 1 value.) ๐ง
okay now the rules, and these ones are serious ๐ญ
When the client side of a connection receives a MODE command, it MUST agree with at least the state of the EDIT and TRAPSIG bits. if a MODE command comes in with a mode mask that's already in use (ignoring the MODE_ACK bit), the MODE command is ignored. if a MODE command comes in that's different from the current mode mask, a reply gets sent with either the new mode mask and the MODE_ACK bit set, or a subset of the new mode mask. the only exception: if the server receives a MODE with either the EDIT or TRAPSIG bits not set, it may set the EDIT and TRAPSIG bits in the response. and if the client receives a MODE with the EDIT or TRAPSIG bits set, it may not clear them in the response ๐
last bit of this part, and it's the glue that keeps everyone in sync ๐ซถ when a MODE command is received with the MODE_ACK bit set, and the mode is different than what the current mode is, the client will ignore the new mode, and the server will switch to the new mode. this makes sure both sides of the connection resolve to the same mode. in all cases, a response is never generated to a MODE command that has the MODE_ACK bit set โจ
2.3 LINEMODE suboption FORWARDMASK ๐
okay bestie, this one is a big ask ๐ซถ The sender is basically saying "please send any buffered data the moment any of the ASCII characters in this bit mask show up." Only the side that sent DO LINEMODE (the server side) may negotiate this, so it's a server-side power move ๐
IAC SB LINEMODE DO FORWARDMASK mask0 mask1 ... mask31 IAC SE
The mask is up to 32 octets long, and each octet stands for 8 ASCII character codes. The high order bit of mask0 is ASCII code 0, and the low order bit of mask0 is ASCII code 7. The high order bit of mask1 is ASCII code 8, and the low order bit of mask1 is ASCII code 15. It keeps going like that from there ๐
The mask list may get cut off before the end. When that happens, all the leftover mask octets are assumed to be reset, which means they're equal to zero ๐ง
When the server side is in DONT TRANSMIT-BINARY mode, only the first 16 octets of the mask get used (ASCII codes 0 through 127) ๐
One more thing: if any individual octet of the mask is equal to IAC, it must be sent as a double IAC โจ
DONT FORWARDMASK ๐ฆ
IAC SB LINEMODE DONT FORWARDMASK IAC SE
The sender of this command is requesting that the other side stop using the forward mask to figure out when to send buffered data. Bye bye, mask ๐
WILL FORWARDMASK ๐ธ
IAC SB LINEMODE WILL FORWARDMASK IAC SE
This one is sent in response to a DO FORWARDMASK command. It means the forward mask will be used to decide when to send buffered data, so the vibes are yes ๐
WONT FORWARDMASK ๐
IAC SB LINEMODE WONT FORWARDMASK IAC SE
This is also sent in response to a DO FORWARDMASK command, but it's the opposite answer. The forward mask will not be used to decide when to send buffered data ๐
2.4 LINEMODE suboption SLC, Set Local Characters ๐
the SLC suboption is a list of octet triplets, and each triplet is tiny but mighty ๐ง The first octet names the function, the second octet holds modifiers for that function, and the third octet is the ASCII character used for the function.
IAC SB LINEMODE SLC <list of octet triplets> IAC SE
The sender of this command REQUESTS that the list of octet triplets be used to set the local character that gets sent to perform the specified function. It's giving "change my settings, please" ๐
Every function can sit at one of four levels. From lowest to highest, they are:
SLC_NOSUPPORT, the lowestSLC_CANTCHANGE, one step upSLC_VALUE, above thatSLC_DEFAULT, the highest
Here's how the second octet's SLC_LEVELBITS work:
- If
SLC_LEVELBITSis equal toSLC_DEFAULT, this function should use the system default on the other side of the connection. - If
SLC_LEVELBITSis equal toSLC_VALUE, the function is supported, and its current value is in the third octet. - If
SLC_LEVELBITSis equal toSLC_CANTCHANGE, the function is supported, but the value in the third octet can't be changed ๐ - If
SLC_LEVELBITSis equal toSLC_NOSUPPORT, the function isn't supported and should be disabled by the other side.
If this is a reply to an earlier request to change a special character, and you're agreeing to the change, then the SLC_ACK bit must be set in the second octet โ
If the SLC_FLUSHIN bit is set in the second octet, then every time this function is sent, a Telnet "sync" should be sent at the same time to flush the input stream ๐งน
If the SLC_FLUSHOUT bit is set in the second octet, then every time this function is sent, output data should be flushed ๐
Only the client may send an octet triplet whose first octet is zero. In that case,SLC_LEVELBITSmay only be set toSLC_DEFAULTorSLC_VALUE, and the third octet doesn't matter. When the server receives0 SLC_DEFAULT 0, it should switch to its system default special character settings and send all of those special characters to the client. When the server receives0 SLC_VALUE 0, it should just send its current special character settings.
Also, if the server doesn't support some of the editing functions, they should be sent as XXX SLC_DEFAULT 0 instead of XXX SLC_NOSUPPORT 0. That way the client can use its own values for those functions and doesn't have to turn them off even when it supports them ๐
If any of the octets in the list of octet triplets is equal to IAC, it must be sent as a double IAC ๐ซ
When a connection is set up, it's the client's job to either request the remote default values for the special characters, or to send over what all the special characters should be set to ๐
The function values fall into two groups: ones that get translated into their Telnet equivalents before being sent across the connection, and ones that get recognized and processed locally. Let's go through them ๐ธ
First up, the characters that get mapped to Telnet equivalents:
SLC_SYNCHSynch. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_BRKBreak. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_IPInterrupt Process. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_AOAbort Output. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_AYTAre You There. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_EOREnd of Record. See RFC 885, "TELNET END OF RECORD OPTION" for a complete description.SLC_ABORTAbort. See section 2.5 for a complete description.SLC_EOFEnd of File. See section 2.5 for a complete description.SLC_SUSPSuspend. See section 2.5 for a complete description.
Next, the locally interpreted functions ๐ง
SLC_ECErase Character. This is the character you type to erase one character from the input stream. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_ELErase Line. This is the character you type to erase the entire contents of the current line of input. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description.SLC_EWErase Word. This is the character you type to erase one word from the input stream.SLC_RPReprint Line. This is the character you type to reprint the current line of input, leaving the cursor at the end of the line ๐SLC_LNEXTLiteral Next. This is the character you type to say the next character should be taken literally. No character processing should be done on it, and if it's a special character that would normally get mapped to a Telnet option, that mapping should not be done. Rude but iconic ๐SLC_XONStart Output. This is the character sent to resume output to the user's terminal โถ๏ธSLC_XOFFStop Output. This is the character sent to stop output to the user's terminal โธ๏ธSLC_FORW1Forwarding character. This is a character that should cause all data currently being buffered, plus this character, to be sent immediately ๐SLC_FORW2Forwarding character. This is another character that is treated the same way asSLC_FORW1๐SLC_MCLMove cursor one character left. When visual editing is supported, this is the character that moves the cursor one character to the left in the display โฌ ๏ธSLC_MCRMove cursor one character right. When visual editing is supported, this is the character that moves the cursor one character to the right in the display โก๏ธSLC_MCWLMove cursor one word left. When visual editing is supported, this is the character that moves the cursor one word to the left in the display โชSLC_MCWRMove cursor one word right. When visual editing is supported, this is the character that moves the cursor one word to the right in the display โฉSLC_MCBOLMove cursor to the beginning of the line. When visual editing is supported, this is the character that moves the cursor to the beginning of the line being edited ๐SLC_MCEOLMove cursor to the end of the line. When visual editing is supported, this is the character that moves the cursor to the end of the line being edited ๐SLC_INSRTEnter insert mode. When visual editing is supported, once this character is typed, all normal characters typed after it get inserted into the display โSLC_OVEREnter overstrike mode. When visual editing is supported, once this character is typed, all normal characters typed after it overwrite whatever is in the current display. IfSLC_INSRTandSLC_OVERare set to the same value, that value acts as a toggle between overstrike and insert mode ๐SLC_ECRErase character to the right. When visual editing is supported, this is the character that erases one character to the right of the cursor ๐งฝSLC_EWRErase word to the right. When visual editing is supported, this is the character that erases one word to the right of the cursor ๐งฝSLC_EBOLErase to the beginning of the line. When visual editing is supported, this is the character that erases all the characters to the left of the cursor ๐งนSLC_EEOLErase to the end of the line. When visual editing is supported, this is the character that erases all characters to the right of the cursor ๐งน
For SLC_EEOL, SLC_EWR, and SLC_ECR, there's a detail to know. If a system has a cursor that isn't displayed between characters but is positioned over a character, that character is assumed to be to the right of the cursor. So SLC_ECR erases the character that's under the current cursor position ๐ฆ
2.5 New control characters โจ
okay bestie, this section adds three brand new control characters so the NVT can do stuff the old spec just couldn't ๐ ๐
IAC ABORT ๐ฆ
Abort. It's similar to IAC IP, but it only means to abort or terminate the process the NVT is connected to. (The Telnet spec says IP may "suspend, interrupt, abort or terminate" the process ๐) If a system doesn't have two methods of interrupting a process, then ABORT and IP SHOULD have the same effect. the logic is literally iconic ๐
IAC SUSP ๐ง
Suspend the execution of the current process attached to the NVT in such a way that another process takes over control of the NVT, and the suspended process can be resumed later. If the receiving system doesn't support this functionality, it SHOULD be ignored. no drama, just vibes ๐ธ
IAC EOF ๐
End Of File. The recipient SHOULD notify the process connected to the NVT that an end of file has been reached. This is meant for systems that let the user type an EOF character on the keyboard, so it's giving keyboard-first energy โจ๏ธ๐
3. Default Specification
The default specification for this option is:
WONT LINEMODE
DONT LINEMODE
which literally means there won't be any subnegotiation of the mode of the connection. quiet girl era ๐คซ
If WILL LINEMODE is negotiated, the defaults are:
IAC SB LINEMODE MODE 0 IAC SE
IAC SB LINEMODE WONT FORWARDMASK IAC SE
If DO LINEMODE is negotiated, the defaults are:
IAC SB LINEMODE MODE 0 IAC SE
IAC SB LINEMODE DONT FORWARDMASK IAC SE
Character values for SLC default to SLC_NOSUPPORT. ๐ซถ
4. Motivation
okay so the big why ๐ฅ As Telnet got more popular, it became super clear that doing command line processing on the local machine and then sending finished lines to the remote machine is a must-have in a bunch of environments. First, over long delay equipment, it's SO frustrating when your typed stuff takes several seconds to echo back ๐ญ Second, some supercomputers aren't great at handling single character input. For those machines it's way better to let the front end computer handle the characters and leave the supercomputer's cycles free for vectorized number crunching ๐งฎโจ
Okay but there were already attempts to make local line editing work inside the existing Telnet specs. The 4.3 BSD tape has a version of Telnet that tries it by recognizing the state of the ECHO and SUPRESS-GO-AHEAD options, and other implementations just do the recognition through the ECHO option alone ๐
Both of these have problems. Using just ECHO gives you no way to turn echoing off for the user while keeping local character processing on, like when someone is typing a password ๐
The usage of SUPRESS-GO-AHEAD comes from reading RFC 858, which says:
"In many TELNET implementations it will be desirable to couple the SUPRESS-GO-AHEAD option to the echo option so that when the echo option is in effect, the SUPPRESS-GO-AHEAD option is in effect simultaneously: both of these options will normally have to be in effect simultaneously to effect what it commonly understood to be character at a time echoing by the remote computer."
The reverse reading of that is that without ECHO or SUPPRESS-GO-AHEAD, you're in line at a time mode, which implies local line editing. The obvious problem is that this isn't what SUPPRESS-GO-AHEAD is supposed to mean ๐ฌ
Other shortcomings: the Telnet spec isn't rich enough to handle all the special characters some current operating systems support. For example, the ECHO/SGA implementation supports two ways of interrupting a process by borrowing the BRK option for the second interrupt. Some implementations have taken the EOR option to send an End-Of-File. Obviously, that's using things for purposes they weren't built for, and the correct solution is to define new options ๐
Another problem is that some line mode implementations buffer up the input until the end of the line, then send the whole line across, editing characters and all. No local editing has actually happened there ๐
After looking at several implementations, it became clear that the right move is to add new options to enhance the current Telnet spec, so it can support local line editing in a reasonable, reliable, and consistent way ๐
There are three states that are of interest:
- Local line editing and local signal trapping
- Remote line editing, local signal trapping
- Remote line editing, remote signal trapping
Local line editing and remote signal trapping isn't really an interesting combo, because you don't recognize the signals and can't send them to the remote side until the line is completed. Also, special signals usually affect line editing, so if they aren't trapped locally, the desired action just won't happen ๐ โโ๏ธ
Local line editing means all the normal command line character processing, like "Erase Character" and "Erase Line", happens on the local system, and only when "CR LF" (or some other special character) shows up is the edited data sent to the remote system ๐
Signal trapping means that if the user types the character associated with the IP function, then the "IAC IP" function is sent to the remote side instead of the typed character. Remote signal trapping means that if the user types that same character, "IAC IP" is NOT sent to the remote side. The actual character typed is sent instead ๐ฆ
5. Implementation Rules
It's expected that any implementation supporting the Telnet LINEMODE option will support all of this specification. no half-stanning allowed ๐
5.1 User Interface
The entire user interface is normally left up to the implementor, but there's some functionality the user should be able to control on the client side of the connection. During a Telnet session, the client side SHOULD offer some mechanism for the user to give commands to the local Telnet process. These commands SHOULD at least let the user:
- Change the mode of the connection. The user SHOULD be able to attempt to turn EDIT, FLOW, TRAPSIG, and ECHO on and off. The server MAY refuse to change the state of the EDIT and TRAPSIG bits. ๐
- Import or export SLC. The user SHOULD be able to tell the local Telnet process whether they want to use the local definitions or the current or default remote definitions of the special characters. ๐ฅ
- Manual sending of options. The user SHOULD be able to tell the local Telnet process to explicitly send any of the Telnet options (like IP, ABORT, AYT, etc). โจ
5.2 End of line terminators.
When LINEMODE is on, and when in EDIT mode, any normal line terminator on the client side's operating system that gets typed SHOULD result in the line being transmitted with CR LF as the line terminator. When EDIT mode is off, a carriage return SHOULD be sent as CR NUL, a line feed SHOULD be sent as LF, and any other key that can't be mapped into an ASCII character but still means the line is complete (like a DOIT or ENTER key) SHOULD be sent as CR LF. ๐
5.3 Output processing
No matter what mode was negotiated, the server side is responsible for all output processing. Specifically, it SHOULD send CR LF when it wants the "newline" function, CR NUL when it wants just a carriage return, and LF when it wants just a linefeed. ๐ธ
5.4 A terminal driver in Telnet?
Conforming implementations don't need to do all the line editing themselves, and that's totally fine ๐ซถ There's nothing wrong with letting the system terminal driver handle the line editing and then hand the Telnet application the completed, edited line, which then gets sent to the remote system. main character energy for the terminal driver ๐
5.5 Setting of Local Characters
When this RFC was being developed, the original plan was that both sides would use their own defaults for the special characters, even if they weren't the same on both sides. But if that scheme is used, the user sees their local special characters being used, and the remote settings don't matter to them at all. So it was decided that the client side of the connection SHOULD be in control of the character settings. the client is the main character, periodt ๐
When LINEMODE is negotiated, the client MUST either export its local character settings to the server, or send a request (SLC 0 SLC_DEFAULT 0) to import the server's special characters. Usually a client on a full fledged computer would export the special characters, and a client on a machine with no local defaults (like some terminal servers) would import them ๐ฅ๏ธโจ
When an SLC command is received, the action taken SHOULD be:
- Ignore it if it is the same as the current settings.
- If the
SLC_LEVELBITSare the same as the current level bits, but the value is different and theSLC_ACKbit is set, no reply is generated. On the server side, the command is ignored, and on the client side, a switch is made to the new value. This is so that if a request to change the same character is generated by both the server and the client, they will both settle on the client's requested value. ๐ค - If we agree with the new setting, we switch to it and reply with the same value, but also set the
SLC_ACKbit. ๐ - If we don't agree, we send a response with what we think the value SHOULD be. The
SLC_ACKbit is NOT set in this case. You MAY only disagree with a value by sending a different value at a lower level. ๐ โโ๏ธ
If the remote system doesn't support some of the line editing characters, but the front end does, then the front end MAY use the local definitions for those characters when in line mode. In this case, the server SHOULD send SLC xxx SLC_DEFAULT 0 in response to a SLC 0 SLC_DEFAULT 0 request, and just ack whatever value the client requests to set the function to. ๐ง
The SLC_FORW2 character SHOULD only be used if SLC_FORW1 is already in use. ๐ฆ
5.6 FORWARDMASK and SLC_FORW1 and SLC_FORW2
okay so the client side gets a little help here ๐ there are two ways to set forwarding characters, so implementing it doesn't have to be a whole nightmare ๐ญ SLC_FORW1 and SLC_FORW2 let you set two extra characters to forward buffered input data on โจ Lots of terminal drivers can already set one or more line delimiters, so it's pretty easy to support these without going through the local terminal driver, which means you don't have to shove a whole terminal driver into Telnet ๐
And if the local terminal driver already has something that maps nicely onto the FORWARDMASK, that's easy to support too. If it doesn't, then supporting FORWARDMASK takes more work ๐
Also, the client side is REQUIRED to forward data when it sees one of SLC_FORW1, SLC_FORW2, or FORWARDMASK characters, or when any normal line termination or special signal is encountered. The client side is also free to forward on other characters that it chooses ๐ฆ For example, say the server side sent a FORWARDMASK asking for data to be forwarded on the first 20 control characters (ASCII codes 1 through 024). If the client side can't make its local terminal driver forward on just those first 20, but it can make the driver forward on any control character (ASCII codes 1 through 039), then the client side could validly accept the FORWARDMASK and forward on any control character. Main character energy ๐
When in EDIT mode, care SHOULD be taken not to forward at random times, because once that data is forwarded, no more editing can be done on the forwarded part of the line ๐ The only time (other than the normal times) that data SHOULD be forwarded in EDIT mode is if a single input line is too long to handle locally.
5.7 Valid and invalid modes and values
bestie, this part is all about what's allowed and what's not ๐ โโ๏ธ
At no time SHOULD DO LINEMODE be negotiated in both directions of the Telnet connection. The side that is the DO LINEMODE side is considered the server side, and the side that is WILL LINEMODE is the client side ๐ง
At no time SHOULD SB LINEMODE DO/DONT FORWARDMASK be sent unless DO LINEMODE has been previously negotiated. At no time SHOULD SB LINEMODE WILL/WONT FORWARDMASK be sent unless WILL LINEMODE has been previously negotiated. Not me trying to send these out of nowhere ๐
If an ABORT, EOF or SUSP is received and the system doesn't support that functionality, it MAY just be ignored. No drama ๐ซถ
5.8 Flushing input and output
okay this section is giving "clean up your mess" energy ๐งน When an IP, BRK or ABORT is sent, it's usually desirable to flush the input stream, and to flush output to the user until the IP, BRK, or ABORT is processed. The SLC_FLUSHIN and SLC_FLUSHOUT bits say what action should happen. These bits are advisory only, but they SHOULD be honored if possible โจ
The standard way to process SLC_FLUSHIN is the Telnet "Synch" signal, and SLC_FLUSHOUT is processed using the TIMING-MARK option. If both are to be sent, the IAC DM goes out before the DO TIMING-MARK. So the sender would send IAC XXX IAC DM IAC DO TIMING-MARK, where XXX may be IP, BRK or ABORT, or any other special character. The IAC DM is sent as TCP urgent data with the DM as the last (or only) data octet, and that's what flushes the input stream ๐ The IAC DO TIMING-MARK tells everyone when to stop flushing output. Once it's sent, all data is discarded until an IAC WILL TIMING-MARK or an IAC WONT TIMING-MARK is received ๐
Since SLC_FLUSHIN and SLC_FLUSHOUT are only advisory, the user interface SHOULD provide a method so the user can override the sending (or not sending) of the "Synch" and TIMING-MARK. The default action, though, SHOULD be to send them according to the SLC_FLUSHIN and SLC_FLUSHOUT bits ๐
Whenever anIAC AOis received, a Synch MUST be returned. Whenever a Synch is being processed (by the TCP connection going into Urgent mode), all data MUST be discarded (but not Telnet commands!) until anIAC DMis found and the connection goes out of Urgent mode. See RFC 854, "TELNET PROTOCOL SPECIFICATION", for a complete description of the Synch signal. Do not skip this one ๐ญ
5.9 State diagram for SLC
okay so this is the state diagram for SLC, and it's kind of a lot ๐ but it's giving organized chaos ๐ read it like a little map and you'll be fine ๐ง
+---------------------------------------------------------------+
| IDLE |
+----------------------+------+------+-------+-------+---------++
^ ^ ^ | | ^ | ^ | ^ | ^ |
| | | v v | | | | | v | |
| | | +------+ +---+--+ | | | | ########### | |
| | | | Get | | Send | | | | | # Get # | |
| | | | SPC0 | | SPC0 | | | | | # 0,DEF,0 # | |
| | | +---+--+ +------+ | | | | ########### | |
| | | | ^ | | | | | | |
| | | v | v | | | v | |
| | | / \ | *********** | | ########### | |
| | | / \ | * Send * | | # Switch # | |
| ********** |Yes/ Same as \ | * 0,VAL,0 * | | # to # | |
| * Change * +--< current? > | *********** | | # default # | |
| * to new * \ / | v | ########### | |
| * value * \ / | *********** | | |
| ********** \ / | * Send * v | |
| ^ |No | * 0,DEF,0 * ######### | |
| |Yes v | *********** # Send #--+ |
| / \ / \ | # SPC-A # |
| / \ / \ | ######### |
| / Is ACK \ Yes/ Same \ | ^ |
|< bit set? ><-< level as > | | |
| \ / \ current?/ | ########### |
| \ / \ / | # Get #<--+
| \ / \ / +-+---+ # 0,VAL,0 #
| |No |No | Set | ###########
| +--------------+ | ACK |
| v | bit | * - Client side only
| / \ +-----+ # - Server side only
| +------+ / \ ^
| | Send | No / Do we \ Yes|
+---| SPC1 |<---< agree? >---+
+------+ \ /
\ /
\ /
SPC0 Initial setting for a special character
SPC1 A changed special character < SPC0
SPC-A All current special character settings
VAL SLC_VALUE level
DEF SLC_DEFAULT level
Levels: DEFAULT, VALUE, CANT_CHANGE, NOSUPPORT
Flags: ACK
Receive Response
------- --------
f,SLC_DEFAULT,x f,SLC_VALUE,v
f,SLC_CANTCHANGE,v
f,SLC_NOSUPPORT,x
f,SLC_VALUE,v f,SLC_ACK|SLC_VALUE,v
f,SLC_CANTCHANGE,w
f,SLC_NOSUPPORT,x
f,SLC_CANTCHANGE,v f,SLC_ACK|SLC_CANTCHANGE,v
f,SLC_NOSUPPORT,x
f,SLC_NOSUPPORT,x f,SLC_ACK|SLC_NOSUPPORT,x
x,SLC_ACK|x,x no response
5.10 Examples of a connection โจ
okay so this is where everything finally clicks ๐ฆ the examples use symbolic names instead of the actual values so they're easier to read bestie ๐ซถ when two or more names are joined by a |, that means the real value is the logical "or" of those names. to keep things clean, the leading IAC and IAC SB sequences and the trailing IAC SE sequences are left out, and the SLC_ prefix is dropped wherever it would normally show up ๐
CLIENT SERVER
DO TOGGLE-FLOW-CONTROL
DO LINEMODE
WILL TOGGLE-FLOW-CONTROL
WILL LINEMODE
[ Subnegotiation may now proceed in both directions. The client
sends of the list of special characters. ]
LINEMODE SLC SYNCH DEFAULT 0 IP
VALUE|FLUSHIN|FLUSHOUT 3 AO
VALUE 15 AYT DEFAULT 0 ABORT
VALUE|FLUSHIN|FLUSHOUT 28 EOF
VALUE 4 SUSP VALUE|FLUSHIN 26 EC
VALUE 127 EL VALUE 21 EW VALUE
23 RP VALUE 18 LNEXT VALUE 22
XON VALUE 17 XOFF VALUE 19
[ Now that linemode is enabled, the server sets the initial mode,
and acknowledges the special characters. ]
LINEMODE MODE EDIT
LINEMODE SLC SYNCH NOSUPPORT 0
IP VALUE|FLUSHIN|FLUSHOUT|ACK 3
AO NOSUPPORT 0 AYT NOSUPPORT 0
ABORT VALUE|FLUSHIN|FLUSHOUT|ACK
28 EOF VALUE|ACK 4 SUSP NOSUP-
PORT 0 EC VALUE|ACK 127 EL
VALUE|ACK 21 EW VALUE|ACK 23 RP
VALUE|ACK 18 LNEXT VALUE|ACK 22
XON VALUE|ACK 17 XOFF VALUE|ACK
19
[ The client gets the mode and ack of the special characters, and
acks the mode and any special characters that the server
changed. ]
LINEMODE MODE EDIT|MODE_ACK
LINEMODE SLC SYNCH NOSUPPORT|ACK
0 AO NOSUPPORT|ACK 0 AYT|ACK
NOSUPPORT 0 SUSP NOSUPPORT|ACK 0
"Login:"
"my_account"
[ Turn off echo to the user. ]
WILL ECHO
DO ECHO
"Password:"
"my_password"
[ Turn back on echo to the user. ]
WONT ECHO
DONT ECHO
[ User does some stuff, and then runs an application that wants
to use single character mode, doing its own echoing of
characters, but keep signal trapping on. ]
WILL ECHO
DO ECHO
LINEMODE MODE TRAPSIG
LINEMODE MODE TRAPSIG|MODE_ACK
[ Application finishes ]
WONT ECHO
DONT ECHO
LINEMODE MODE EDIT|TRAPSIG
LINEMODE MODE
EDIT|TRAPSIG|MODE_ACK
[ Another application, that wants full control of everything. ]
WILL ECHO
DO ECHO
LINEMODE MODE 0
LINEMODE MODE 0|MODE_ACK
[ Application finishes ]
WONT ECHO
DONT ECHO
LINEMODE MODE EDIT|TRAPSIG
LINEMODE MODE
EDIT|TRAPSIG|MODE_ACK
[ The user changes his erase character to ^H. ]
LINEMODE SLC EC VALUE 8
LINEMODE SLC EC VALUE|ACK 8
[ The user decides to revert to all the original client side
special characters. ]
LINEMODE SLC SYNCH DEFAULT 0 IP
VALUE|FLUSHIN|FLUSHOUT 3 AO
VALUE 15 AYT DEFAULT 0 ABORT
VALUE|FLUSHIN|FLUSHOUT 28 EOF
VALUE 4 SUSP VALUE|FLUSHIN 26 EC
VALUE 127 EL VALUE 21 EW VALUE
23 RP VALUE 18 LNEXT VALUE 22
XON VALUE 17 XOFF VALUE 19
LINEMODE SLC SYNCH NOSUPPORT 0
AO NOSUPPORT 15 AYT NOSUPPORT 0
SUSP NOSUPPORT|FLUSHIN 26 EC
VALUE|ACK 127 EW VALUE|ACK 23 RP
VALUE|ACK 18 LNEXT VALUE|ACK 22
XON VALUE|ACK 17 XOFF VALUE|ACK
19
LINEMODE SLC SYNCH NOSUPPORT|ACK
0 AO NOSUPPORT|ACK 15 AYT
NOSUPPORT|ACK 0 SUSP
NOSUPPORT|ACK|FLUSHIN 26
[ The user decides to import the remote sides default special
characters. ]
LINEMODE SLC 0 DEFAULT 0
LINEMODE SLC IP
VALUE|FLUSHIN|FLUSHOUT 3 ABORT
VALUE|FLUSHIN|FLUSHOUT 28 EOF
VALUE 4 EC VALUE 127 EL VALUE 21
[ Since these are the same as the current local settings, no
response is generated. ]
[ This next example is what would happen if an editor was fired
up, that wanted to let the client side do the echoing and
buffering of characters, but did not want it to do any line
editing, and only forward the data when got a control
character. Note that we have preceded all the the 0377s
in the forward mask with an IAC. ]
LINEMODE MODE 0
LINEMODE DO FORWARDMASK IAC 0377
IAC 0377 IAC 0377 IAC 0377 0 0 0
0 0 0 0 0 0 0 0 01
LINEMODE MODE 0
LINEMODE WILL FORWARDMASK
[ Application runs to completion, and then things are to be set
back to what they were before. ]
LINEMODE MODE EDIT|TRAPSIG
LINEMODE DONT FORWARDMASK
LINEMODE MODE EDIT|TRAPSIG
LINEMODE WONT FORWARDMASK
6. Other Telnet options and RFCs โจ
okay, here's the reading list ๐ these are the RFCs for the Telnet options that should be supported along with LINEMODE:
- Postel, J. and Reynolds, J., "TELNET PROTOCOL SPECIFICATION", RFC 854, ISI, May 1983.
- Postel, J. and Reynolds, J., "TELNET OPTION SPECIFICATIONS", RFC 855, ISI, May 1983.
- Postel, J. and Reynolds, J., "TELNET BINARY TRANSMISSION", RFC 856, ISI, May 1983.
- Postel, J. and Reynolds, J., "TELNET ECHO OPTION", RFC 857, ISI, May 1983.
- Postel, J. and Reynolds, J., "TELNET SUPRESS GO AHEAD OPTION", RFC 858, ISI, May 1983.
- Postel, J. and Reynolds, J., "TELNET TIMING MARK OPTION", RFC 860, ISI, May 1983.
- VanBokkeln, J., "Telnet Terminal-Type Option", RFC 1091, FTP Software, Inc., February 1989.
- Waitzman, D., "Telnet Window Size Option", RFC 1073, BBN STC, October 1988.
- Hedrick, C., "Telnet Remote Flow Control Option", RFC 1080, Rutgers University, November, 1988.
- Hedrick, C., "Telnet Terminal Speed Option", RFC 1079, Rutgers University, December, 1988.
and then there's the second group ๐ธ these ones don't need to be supported for LINEMODE, but they would seriously level up any TELNET implementation ๐ง
- Postel, J. and Reynolds, J., "TELNET STATUS OPTION", RFC 859, ISI, May 1983.
- Postel, J. and Reynolds, J., "TELNET END OF RECORD OPTION", RFC 885, ISI, December 1983.
- Silverman, S., "OUTPUT MARKING TELNET OPTION", RFC 933, MITRE-Washington, January 1985.
- Marcy, G., "Telnet X Display Location Option", RFC 1096, Carnegie Mellon University, March 1989.
Security Consideratiions ๐
Security issues are not discussed in this memo. yep, that's the whole section, so no security promises coming from this doc ๐
Author's Address ๐
here's where you can reach the author and the working group ๐
David A. Borman
Cray Research Inc.
655F Lone Oak Drive
Eagan, MN 55123
Phone: (612) 452-6650
EMail: dab@CRAY.COM
IETF Telnet WG Mailing List: telnet-ietf@CRAY.COM
๐ฆ where this page comes from
- ietf: rfc1184 1990-10
- ๐ girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms