RFC 1196
Finger User Information Protocol
RFC 1196: The Finger User Information Protocol ๐
Okay bestie, this is RFC 1196 ๐ง It's from the Network Working Group, with request for comments number 1196 ๐. D. Zimmerman from the Center for Discrete Mathematics and Theoretical Computer Science wrote it, and it's dated December 1990 ๐. It's an IAB standards track protocol, and it obsoletes RFCs 1194 and 742 ๐ so those older ones are officially retired now.
Status of this Memo
So this memo is all about defining a protocol for swapping user info around ๐ It's an IAB standards track protocol for the Internet community, and they're literally asking everyone for discussion and suggestions for improvements ๐ฌ. If you want the standardization state and status of this protocol, check the current edition of the "IAB Official Protocol Standards" ๐. And distribution of this memo is unlimited ๐.
Abstract
So this memo describes the Finger User Information Protocol ๐ฆ It's a simple protocol that gives you an interface to a remote user information program, which is giving main character energy โจ. Based on RFC 742, the description of the original Finger protocol, this memo tries to clarify how the two ends of a Finger connection are supposed to talk ๐ฌ. It also tries really hard not to break the many existing implementations or add pointless restrictions to the original protocol definition ๐ซถ. This edition gives RFC 1194 a minor correction and clarification ๐ .
Table of Contents
1. Introduction ........................................... 2
1.1. Intent ............................................... 2
1.2. History .............................................. 3
1.3. Requirements ......................................... 3
2. Use of the protocol .................................... 3
2.1. Flow of events ....................................... 3
2.2. Data format .......................................... 4
2.3. Query specifications ................................. 4
2.4. RUIP {Q2} behavior ................................... 4
2.5. Expected RUIP response ............................... 5
2.5.1. {C} query .......................................... 5
2.5.2. {U}{C} query ....................................... 6
2.5.3. {U} ambiguity ...................................... 6
2.5.4. /W query token ..................................... 6
2.5.5. Vending machines ................................... 7
3. Security ............................................... 7
3.1. Implementation security .............................. 7
3.2. RUIP security ........................................ 7
3.2.1. {Q2} refusal ....................................... 7
3.2.2. {C} refusal ........................................ 8
3.2.3. Atomic discharge ................................... 8
3.2.4. User information files ............................. 8
3.2.5. Execution of user programs ......................... 9
3.2.6. {U} ambiguity ...................................... 9
3.2.7. Audit trails ....................................... 9
3.3. Client security ...................................... 9
4. Examples ............................................... 10
4.1. Example with a null command line ({C}) ............... 10
4.2. Example with name specified ({U}{C}) ................. 10
4.3. Example with ambiguous name specified ({U}{C}) ....... 11
4.4. Example of query type {Q2} ({U}{H}{H}{C}) ............ 11
5. Acknowledgments ........................................ 12
6. Security Considerations ................................ 12
7. Author's Address ....................................... 12
1. Introduction โจ
1.1. Intent
Okay so this memo describes the Finger User Information Protocol ๐ฆ It's a simple protocol that gives you an interface to a remote user information program (RUIP) ๐ .
Based on RFC 742, the description of the original Finger protocol, this memo tries to clarify how the two ends of a Finger connection are supposed to communicate ๐ซถ. It also tries not to break the many current implementations or add unnecessary restrictions to the original protocol definition ๐ธ.
The most popular versions of Finger today seem to mostly come from the BSD UNIX work at the University of California, Berkeley ๐. So this memo is built around how the BSD version behaves โจ.
But here's the tea โ: the BSD version gives you very few options to customize the Finger RUIP for a particular site's security policy, or to protect the user from dangerous data ๐. There are MANY potential security holes that implementors and administrators need to be aware of ๐จ. This is extra sensitive because the whole point of this protocol is to return information about a system's users, which is a sensitive issue at best ๐. So this memo makes a bunch of important security comments and recommendations ๐.
1.2. History
The FINGER program at SAIL, written by Les Earnest, inspired the NAME program on ITS ๐ง. Earl Killian at MIT and Brian Harvey at SAIL were jointly responsible for implementing the original protocol ๐.
Ken Harrenstien is the author of RFC 742, "Name/Finger", which is where this memo started out ๐.
1.3. Requirements
Okay so in this doc, the words that define how important each requirement is get written in capitals ๐ฃ Here's the lineup:
- "MUST": this word, or the adjective "REQUIRED", means the item is an absolute requirement of the specification. No wiggle room ๐
- "SHOULD": this word, or the adjective "RECOMMENDED", means there might be valid reasons in particular circumstances to ignore the item. But you need to understand the full implications and weigh the case carefully before going a different way ๐ค
- "MAY": this word, or the adjective "OPTIONAL", means the item is truly optional. One vendor might include it because a particular marketplace needs it or because it makes the product better, for example. Another vendor might leave the same item out ๐ฌ
An implementation is not compliant if it fails to satisfy one or more of the MUST requirements. An implementation that satisfies all the MUST and all the SHOULD requirements is said to be "unconditionally compliant"; one that satisfies all the MUST requirements but not all the SHOULD requirements is said to be "conditionally compliant" ๐ฏ.
2. Use of the protocol ๐ฎ
2.1. Flow of events
Finger is built on the Transmission Control Protocol, using TCP port 79 decimal (117 octal) ๐. You open a TCP connection to a remote host on the Finger port, and an RUIP shows up on the far end to handle your request ๐ฌ. The RUIP gets a one-line query based on the Finger query specification. Then it processes the query, sends back an answer, and closes the connection normally ๐.
2.2. Data format
Any data transferred MUST be in ASCII format, with no parity, and with lines ending in CRLF (ASCII 13 followed by ASCII 10) ๐. This rules out other character formats such as EBCDIC, etc ๐ซ. It also means that any characters between ASCII 128 and ASCII 255 should truly be international data, not 7-bit ASCII with the parity bit set ๐.
2.3. Query specifications
An RUIP MUST accept the entire Finger query specification ๐ฏ.
The Finger query specification is defined like this ๐:
{Q1} ::= [{U}] [/W] {C}
{Q2} ::= [{U}]{H} [/W] {C}
{U} ::= username
{H} ::= @hostname | @hostname{H}
{C} ::= <CRLF>
{H}, being recursive, means there's no arbitrary limit on how many @hostname tokens a query can have ๐. In examples of the {Q2} request specification, the number of @hostname tokens is limited to two, just to keep things short ๐.
Be aware that {Q1} and {Q2} don't refer to a user typing finger user@host at an operating system prompt ๐. They refer to the line that an RUIP actually receives. So if a user types finger user@host<CRLF>, the RUIP on the remote host receives user<CRLF>, which corresponds to {Q1} ๐ง .
Like everything in the IP protocol suite, "be liberal in what you accept" โจ.
2.4. RUIP {Q2} behavior
A query of {Q2} is a request to forward a query to another RUIP. An RUIP MUST either provide or actively refuse this forwarding service (see section 3.2.1). If an RUIP provides this service, it MUST conform to the following behavior ๐:
Here's the setup:
- Host
<H1>opens a Finger connection<F1-2>to an RUIP on host<H2>. <H1>gives the<H2>RUIP a query<Q1-2>of type{Q2}(e.g.,FOO@HOST1@HOST2).
From that, you can work out that:
- Host
<H3>is the right-most host in<Q1-2>(i.e.,HOST2). - Query
<Q2-3>is the remainder of<Q1-2>after removing the right-most "@hostname" token in the query (i.e.,FOO@HOST1).
And that means:
- The
<H2>RUIP then must itself open a Finger connection<F2-3>to<H3>, using<Q2-3>. - The
<H2>RUIP must return any information received from<F2-3>to<H1>via<F1-2>. - The
<H2>RUIP must close<F1-2>in normal circumstances only when the<H3>RUIP closes<F2-3>.
2.5. Expected RUIP response
For the most part, an RUIP's output doesn't follow a strict specification, since it's meant to be read by people instead of programs ๐งโ๐ป. It should mainly strive to be informative โจ.
Output of ANY query is subject to the discussion in the security section ๐.
2.5.1. {C} query
A query of {C} is a request for a list of all online users ๐ฅ. An RUIP MUST either answer or actively refuse (see section 3.2.2). If it answers, then it MUST provide at least the user's full name ๐. The system administrator SHOULD be allowed to include other useful information (per section 3.2.3), such as:
- terminal location
- office location
- office phone number
- job name
- idle time (number of minutes since last typed input, or since last job activity).
2.5.2. {U}{C} query
A query of {U}{C} is a request for in-depth status of a specified user {U} ๐. If you really want to refuse this service, you probably don't want to be running Finger in the first place ๐
.
An answer MUST include at least the full name of the user. If the user is logged in, at least the same amount of information returned by {C} for that user MUST also be returned by {U}{C} ๐.
Since this is a query for info on one specific user, the system administrator SHOULD be allowed to choose to return additional useful information (per section 3.2.3), such as:
- office location
- office phone number
- home phone number
- status of login (not logged in, logout time, etc)
- user information file
A user information file is a feature where a user can leave a short message that gets included in the response to Finger requests ๐ (This is sometimes called a "plan" file.) It's pretty easy to build this, for example by having the program look for a specially named text file in the user's home directory or some common area. The exact method is up to the implementor ๐ ๏ธ. The system administrator SHOULD be allowed to specifically turn this feature on and off. See section 3.2.4 for caveats โ ๏ธ.
There MAY be a way for the user to run a program in response to a Finger query ๐คฏ. If this feature exists, the system administrator SHOULD be allowed to specifically turn it on and off. See section 3.2.5 for caveats ๐จ.
2.5.3. {U} ambiguity ๐
Okay so the command line is allowed to have "names" in it, and those MUST include "user names" or "login names" as the system defines them ๐ซถ. But what if a name is ambiguous?? Then the system administrator SHOULD be allowed to choose whether or not all possible derivations should be returned in some fashion (per section 3.2.6) โจ.
2.5.4. /W query token ๐
So the /W token in a {Q1} or {Q2} query SHOULD at best be interpreted at the last RUIP to signify a higher level of verbosity in the user information output, or at worst be ignored ๐ธ. Either way, it's giving flexible.
2.5.5. Vending machines ๐
Vending machines SHOULD respond to a {C} request with a list of all items currently available for purchase and possible consumption ๐. When a {U}{C} request comes in, vending machines SHOULD respond with a detailed count or list of the particular product or product slot ๐ง. But vending machines should NEVER NEVER EVER eat requests. Or money. ๐
3. Security ๐ก๏ธ
3.1. Implementation security
Sound implementation of Finger is of the utmost importance, no cap ๐ . Implementations should be tested against various forms of attack. In particular, an RUIP SHOULD protect itself against malformed inputs. Vendors providing Finger with the operating system or network software should subject their implementations to penetration testing.
Finger is one of the avenues for direct penetration, as the Morris worm pointed out quite vividly ๐. Like Telnet, FTP and SMTP, Finger is one of the protocols at the security perimeter of a host. Accordingly, the soundness of the implementation is paramount. The implementation should receive just as much security scrutiny during design, implementation, and testing as Telnet, FTP, or SMTP ๐ฅ.
3.2. RUIP security
Warning!! Finger discloses information about users; moreover, such information may be considered sensitive ๐ญ.
Security administrators should make explicit decisions about whether to run Finger and what information should be provided in responses. One existing implementation provides the time the user last logged in, the time he last read mail, whether unread mail was waiting for him, and who the most recent unread mail was from! ๐ This makes it possible to track conversations in progress and see where someone's attention was focused. Sites that are information-security conscious should not run Finger without an explicit understanding of how much information it is giving away ๐.
3.2.1. {Q2} refusal ๐ซ
For individual site security concerns, the system administrator SHOULD be given an option to individually turn on or off RUIP processing of {Q2}. If RUIP processing of {Q2} is turned off, the RUIP MUST return a service refusal message of some sort. "Finger forwarding service denied" is adequate. The purpose of this is to allow individual hosts to choose to not forward Finger requests, but if they do choose to, to do so consistently.
Overall, there are few cases which would warrant processing of {Q2} at all, and they are far outweighed by the number of cases for refusing to process {Q2}. In particular, be aware that if a machine is part of security perimeter (that is, it is a gateway from the outside world to some set of interior machines), then turning {Q2} on provides a path through that security perimeter. Therefore, it is RECOMMENDED that the default of the {Q2} processing option be to refuse processing. It certainly should not be enabled in gateway machines without careful consideration of the security implications ๐
โโ๏ธ.
3.2.2. {C} refusal ๐ซ
For individual site security concerns, the system administrator SHOULD be given an option to individually turn on or off RUIP acceptance of {C}. If RUIP processing of {C} is turned off, the RUIP MUST return a service refusal message of some sort. "Finger online user list denied" is adequate. The purpose of this is to allow individual hosts to choose to not list the users currently online.
3.2.3. Atomic discharge โก
All implementations of Finger SHOULD allow individual system administrators to tailor what atoms of information are returned to a query. For example:
- Administrator A should be allowed to specifically choose to return office location, office phone number, home phone number, and logged in/logout time.
- Administrator B should be allowed to specifically choose to return only office location, and office phone number.
- Administrator C should be allowed to specifically choose to return the minimum amount of required information, which is the person's full name.
3.2.4. User information files ๐
Allowing an RUIP to return information out of a user-modifiable file should be seen as equivalent to allowing any information about your system to be freely distributed. That is, it is potentially the same as turning on all specifiable options. This information security breach can be done in a number of ways, some cleverly, others straightforwardly. This should disturb the sleep of system administrators who wish to control the returned information ๐.
3.2.5. Execution of user programs ๐โโ๏ธ
Allowing an RUIP to run a user program in response to a Finger query is potentially dangerous. BE CAREFUL!! -- the RUIP MUST NOT allow system security to be compromised by that program. Implementing this feature may be more trouble than it is worth, since there are always bugs in operating systems, which could be exploited via this type of mechanism ๐.
3.2.6. {U} ambiguity ๐
Be aware that a malicious user's clever and/or persistent use of this feature can result in a list of most of the usernames on a system. Refusal of {U} ambiguity should be considered in the same vein as refusal of {C} requests (see section 3.2.2).
3.2.7. Audit trails ๐
Implementations SHOULD allow system administrators to log Finger queries.
3.3. Client security ๐งผ
It is expected that there will normally be some client program that the user runs to query the initial RUIP. By default, this program SHOULD filter any unprintable data, leaving only printable 7-bit characters (ASCII 32 through ASCII 126), tabs (ASCII 9), and CRLFs. This is to protect against people playing with terminal escape codes, changing other peoples' X window names, or committing other dastardly or confusing deeds ๐ฆ. Two separate user options SHOULD be considered to modify this behavior, so that users may choose to view international or control characters:
- one to allow all characters less than ASCII 32
- another to allow all characters greater than ASCII 126
For environments that live and breathe international data, the system administrator SHOULD be given a mechanism to enable the latter option by default for all users on a particular system. This can be done via a global environment variable or similar mechanism ๐.
4. Examples ๐
4.1. Example with a null command line ({C})
Site: elbereth.rutgers.edu
Command line: <CRLF>
Login Name TTY Idle When Office
rinehart Mark J. Rinehart p0 1:11 Mon 12:15 019 Hill x3166
greenfie Stephen J. Greenfiel p1 Mon 15:46 542 Hill x3074
rapatel Rocky - Rakesh Patel p3 4d Thu 00:58 028 Hill x2287
pleasant Mel Pleasant p4 3d Thu 21:32 019 Hill 908-932-
dphillip Dave Phillips p5 021: Sun 18:24 265 Hill x3792
dmk David Katinsky p6 2d Thu 14:11 028 Hill x2492
cherniss Cary Cherniss p7 5 Mon 15:42 127 Psychol x2008
harnaga Doug Harnaga p8 2:01 Mon 10:15 055 Hill x2351
brisco Thomas P. Brisco pe 2:09 Mon 13:37 h055 x2351
laidlaw Angus Laidlaw q0 1:55 Mon 11:26 E313C 648-5592
cje Chris Jarocha-Ernst q1 8 Mon 13:43 259 Hill x2413
4.2. Example with name specified ({U}{C})
Site: dimacs.rutgers.edu
Command line: pirmann<CRLF>
Login name: pirmann In real life: David Pirmann
Office: 016 Hill, x2443 Home phone: 989-8482
Directory: /dimacs/u1/pirmann Shell: /bin/tcsh
Last login Sat Jun 23 10:47 on ttyp0 from romulus.rutgers.
No unread mail
Project:
Plan:
Work Schedule, Summer 1990
Rutgers LCSR Operations, 908-932-2443
Monday 5pm - 12am
Tuesday 5pm - 12am
Wednesday 9am - 5pm
Thursday 9am - 5pm
Saturday 9am - 5pm
larf larf hoo hoo
4.3. Example with ambiguous name specified ({U}{C})
Site: elbereth.rutgers.edu
Command line: ron<CRLF>
Login name: spinner In real life: Ron Spinner
Office: Ops Cubby, x2443 Home phone: 463-7358
Directory: /u1/spinner Shell: /bin/tcsh
Last login Mon May 7 16:38 on ttyq7
Plan:
ught i
ca n
m a
' ... t
I . . i
! m
! ! e
p !pool
l
e
H
Login name: surak In real life: Ron Surak
Office: 000 OMB Dou, x9256
Directory: /u2/surak Shell: /bin/tcsh
Last login Fri Jul 27 09:55 on ttyq3
No Plan.
Login name: etter In real life: Ron Etter
Directory: /u2/etter Shell: /bin/tcsh
Never logged in.
No Plan.
4.4. Example of query type {Q2} ({U}{H}{H}{C})
Site: dimacs.rutgers.edu
Command line: hedrick@math.rutgers.edu@pilot.njin.net<CRLF>
[pilot.njin.net]
[math.rutgers.edu]
Login name: hedrick In real life: Charles Hedrick
Office: 484 Hill, x3088
Directory: /math/u2/hedrick Shell: /bin/tcsh
Last login Sun Jun 24 00:08 on ttyp1 from monster-gw.rutge
No unread mail
No Plan.
5. Acknowledgments ๐
Thanks to everyone in the Internet Engineering Task Force for their comments. Special thanks to Steve Crocker for his security recommendations and prose ๐ซถ.
6. Security Considerations ๐
Security issues are discussed in Section 3.
7. Author's Address ๐๏ธ
David Paul Zimmerman Center for Discrete Mathematics and Theoretical Computer Science Rutgers University P.O. Box 1179 Piscataway, NJ 08855-1179
Phone: (908)932-4592
EMail: dpz@dimacs.rutgers.edu
๐ฆ where this page comes from
- ietf: rfc1196 1990-12
- ๐ girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms