Internet Engineering Task Force Jacob Palme
INTERNET-DRAFT Stockholm University
Expires: December, 2005 Royal Institute of Technology
Informational 7 June 2005
The IETF Golden Rules
draft-palme-golden-rules-00.txt
1. Abstract
This memo presents the following rules, which to some extend can be
regarded as the golden rules of IETF, even though there are exceptions
when these rules should not be adhered to.
- Be liberal in what you accept, and conservative in what you send
- Do not munge forwarded data
- Modify as late as possible
- Cause no harm
- Leave nothing undefined
- Keep it simple, stupid
- No voting, rough consensus
- Plain ASCII text is enough
2. IPR Statement
By submitting this Internet-Draft, each author represents
that any applicable patent or other IPR claims of which he
or she is aware have been or will be disclosed, and any of
which he or she becomes aware will be disclosed, in
accordance with Section 6 of BCP 79.
This document is subject to the rights, licenses
and restrictions contained in BCP 78, and except as set forth
therein, the authors retain all their rights.
Copyright (C) The Internet Society (2005).
This document and the information contained herein
are provided on an "AS IS" basis and THE CONTRIBUTOR, THE
ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE
INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM
ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT
INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY
OR FITNESS FOR A PARTICULAR PURPOSE.
Internet-Drafts are working documents of the Internet
Engineering Task Force (IETF), its areas, and its
working groups. Note that other groups may also distribute
working documents as Internet-Drafts. Internet-Drafts
are draft documents valid for a maximum of six months and
may be updated, replaced, or obsoleted by other documents
at any time. It is inappropriate to use Internet-Drafts as
reference material or to cite them other than as "work in
progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/1id-abstracts.html
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html
This Internet-Draft will expire on 7 December 2005.
3. Table of contents
1. Abstract 1
2. IPR Statement 1
3. Table of contents 2
4. Mailing list and web area 3
5. Introduction 3
6. Rule one: Be liberal in what you accept, and conservative in what
you send 3
7. Rule two: Do not munge forwarded data 6
8. Rule three: Modify as late as possible 7
9. Rule four: Leave nothing undefined 8
10. Rule five: Cause no harm 8
11. Rule six: Keep it simple, stupid 8
12. Rule seven: No voting, rough consensus 9
13. Rule eight: Plain ASCII text is enough 10
14. Contradiction between the rules 11
15. Other documents 11
16. References 12
17. Acknowledgements 12
18. Author's address 12
19. Intellectual Property Statement 12
20. Full Copyright Statement 13
4. Mailing list and web area
Comments and discussion on this MEMO can be sent to the
ietf-golden@dsv.su.se mailing list. More information about this list,
as well as information how to subscribe, unsubscribe and view the
archives of the list, can be found at
http://lists.dsv.su.se/cgi-bin/mailman/listinfo/ietf-golden
The latest version of this memo can be found at
http://dsv.su.se/jpalme/ietf/golden-rule/.
5. Introduction
The Internet Engineering Task Force (IETF) is the largest and most
successful standards organization in the area of Internet protocols.
The success of IETF, as compared to other standards making
organization, depends on its written and unwritten rules. This memo
summarizes some of the most important of these rules. The rules are not
carved in stone, there are cases where the IETF itself has chosen to
partially break these rules. Still, some of these rules are important
and need to be written down.
This memo presents the rules in a rough priority order, with the most
important rule first. For each rule, some explanation is given, as well
as some examples of how it has been used.
6. Rule one: Be liberal in what you accept,
and conservative in what you send
This is the oldest, most wellknown and most important of the rules. It
is usually attributed to [RFC1123]. Below is a direct quote of the text
in [RFC1123] which explains this rule:
At every layer of the protocols, there is a general rule whose
application can lead to enormous benefits in robustness and
interoperability:
"Be liberal in what you accept, and
conservative in what you send"
Software should be written to deal with every conceivable
error, no matter how unlikely; sooner or later a packet will
come in with that particular combination of errors and
attributes, and unless the software is prepared, chaos can
ensue. In general, it is best to assume that the network is
filled with malevolent entities that will send in packets
designed to have the worst possible effect. This assumption
will lead to suitable protective design, although the most
serious problems in the Internet have been caused by
unenvisaged mechanisms triggered by low-probability events;
mere human malice would never have taken so devious a course!
Adaptability to change must be designed into all levels of
Internet host software. As a simple example, consider a
protocol specification that contains an enumeration of values
for a particular header field -- e.g., a type field, a port
number, or an error code; this enumeration must be assumed to
be incomplete. Thus, if a protocol specification defines four
possible error codes, the software must not break when a fifth
code shows up. An undefined code might be logged (see below),
but it must not cause a failure.
The second part of the principle is almost as important:
software on other hosts may contain deficiencies that make it
unwise to exploit legal but obscure protocol features. It is
unwise to stray far from the obvious and simple, lest untoward
effects result elsewhere. A corollary of this is "watch out
for misbehaving hosts"; host software should be prepared, not
just to survive other misbehaving hosts, but also to cooperate
to limit the amount of disruption such hosts can cause to the
shared communication facility.
One could interpret this rule to mean that there are actually three
versions of each standard:
(1) The actual standard as specified in the standards specification.
(2) What good implementers should accept, and which is a superset of (1).
(3) What good implementers should produce, and which is a subset of (1).
This is shown in this figure:
+-------------------------------------------+
| (2) What you should accept |
| +-------------------------------------+ |
| | (1) The actual standard | |
| | +-------------------------------+ | |
| | | (3) What you should produce | | |
| | +-------------------------------+ | |
| +-------------------------------------+ |
+-------------------------------------------+
But if there are in reality three standards, would it not be better if
the standards document described all three of them? Well, this would
make the standards documents much larger. However, it is more and more
common that standards especially specify, at least in some cases, the
difference between each standard, for example by saying that certain
functions are deprecated or obsolete. Functions marked as deprecated or
obsolete belong to category (2) above, what you should accept but not
produce.
Example, quote from [RFC2068], the HTTP standard:
The "charset" parameter is used with some media types to
define the character set (section 3.4) of the data. When no
explicit charset parameter is provided by the sender, media
subtypes of the "text" type are defined to have a default
charset value of "ISO-8859-1" when received via HTTP. Data in
character sets other than "ISO-8859-1" or its subsets MUST be
labeled with an appropriate charset value.
Some HTTP/1.0 software has interpreted a Content-Type header
without charset parameter incorrectly to mean "recipient
should guess." Senders wishing to defeat this behavior MAY
include a charset parameter even when the charset is
ISO-8859-1 and SHOULD do so when it is known that it will not
confuse the recipient.
This text thus says that it is allowed to send a text with HTTP without
a character set indication, but that it MAY be good to include a
character set indication, even when it has the default value
ISO-8859-1. A good (conservative) implementation should thus always
indicate the character set, even though it is not mandatory.
It is unfortunate that the HTTP standard says that when no character
set is specified, the default is ISO-8859-1, while the e-mail standards
say that when no character set is specified, the default is plain
ASCII. This discrepancy between two mayor standards has caused lots of
confusion and incorrect implementations. The robustness rule quoted
above says that it is good practice to always identify the character
set of HTTP-transmitted text, even though this is formally optional.
The e-mail standard [RFC2046] says:
The default character set, US-ASCII, has been the subject of
some confusion and ambiguity in the past. Not only were there
some ambiguities in the definition, there have been wide
variations in practice. In order to eliminate such ambiguity
and variations in the future, it is strongly recommended that
new user agents explicitly specify a character set as a media
type parameter in the Content-Type header field. "US-ASCII"
does not indicate an arbitrary 7-bit character set, but
specifies that all octets in the body must be interpreted as
characters according to the US-ASCII character set. National
and application-oriented versions of ISO 646 [ISO-646] are
usually NOT identical to US-ASCII, and in that case their use
in Internet mail is explicitly discouraged. The omission of
the ISO 646 character set from this document is deliberate in
this regard. The character set name of "US-ASCII" explicitly
refers to the character set defined in ANSI X3.4-1986
[US-ASCII]. The new international reference version (IRV) of
the 1991 edition of ISO 646 is identical to US-ASCII. The
character set name "ASCII" is reserved and must not be used
for any purpose.
Note that again, the standard says that the Character-set attribute is
not mandatory, but good (conservative) implementations should always
specify it for textual body parts in e-mail. Both e-mail and HTTP
standards does say that this attribute is optional but should be used
in what you produce.
One of the most comprehensive explicit specification of the difference
between the recommended and the allowed standard is the e-mail body
format standard, [RFC2822]. This standard has a special chapter 4, with
the title "Obsolete syntax". This chapter goes into great detail in
specifying everything a good implementation should accept, but not
produce. Probably, such explicit specification of what a good
implementation should accept but not produce will occur in the future.
When a standard grows old, the experience on how to interpret the
liberal-conservative rule increases.
Note: By "produce" in Rule one is meant new data which an agent
produces. It does normally not apply to data which an agent receives
and forwards. For such data, instead, the "Rule two: Do not munge
forwarded data" below applies.
7. Rule two: Do not munge forwarded data
It may seem natural for an implementer, when receiving incorrect data,
to correct it before sending it further. Such corrections of incorrect
data is in IETF vocabulary called "munging", and it is in general
disapproved. The reason for this is that long experience has shown that
attempts to correct incorrect data will usually cause more confusion
than keeping it as it is. This might be seen as a contraction to the
liberal-conservative rule, but it is not, since the
liberal-conservative rule talks about what you produce, not about what
you forward.
A commonly occuring case where implementers wrongly do not adhere to
"Rule two: Do not munge" is certain mail systems which when it receives
a message converts it to an internal format. When the same message is
then sent out again, it is converted back again to standard format. The
correct behaviour of such systems is that even if they need to convert
the message to an internal format to be able to handle it locally, they
should keep a copy of the original message before conversion, and use
this when sending it out again. A much-used mailer will for example
change the Message-ID when resending a message. Doing this is bad,
because it can cause mailing-list loops, because threads are broken and
for other reasons.
There are of course cases where the "do not munge" rule should not be
adhered to. A well-known example is certain attributes in submission to
Mail Transfer Agents (MTAs). Some MTAs will correct incorrect mail by
adding missing "Date:" and "Content-ID" headers. This is in general not
good. For example, if an MTA adds "Content-ID" to a message which it
receives by SMTP from another MTA, different versions of the same
message may turn up with different values of the Content-ID, which can
cause various problems. However, the first MTA which receives a message
from the originating Message User Agent (MUA) can add a missing
Content-ID without this risk. Therefore, a specific standard has been
written, [RFC2476] which specifies what kind of munging is allowed and
not allowed for the first MTA which receives a message from an MUA.
There are certain advantages with doing this in some cases, for
example, the submission MTA may have a more reliable clock than the
originating MUA.
8. Rule three: Modify as late as possible
If you absolutely have to modify incoming data, this modification
should be done as late as possible. If possible, standards should be
written so that data need not be modified. The reason for this rule is
that modifying or converting data often causes some damage to the data
(either loss of information or addition of possibly incorrect
information or both). And if this conversion is done as late as
possible, it is easier for the recipient to control how it is done. For
example, if you receive an HTML document which your browser cannot
display properly, you can choose to display it using another web
browser instead of converting it. Another reason is that you may need
to send out what you have received. And when re-sending information,
"Rule two: Do not munge" usually means that the original document
should be sent, unless you explicitly under manual control make changes
to it or convert it.
This is the reason, for example, for the use of the Content-Location
heading in the MHTML standard [MHTML]. Using Content-Location, HTML
documents can be fetched unchanged from a web site and put into e-mail.
The alternative (Content-ID, which is also allowed according to MHTML)
usually requires rewriting of the HTML code when it is to be sent by e-
mail.
9. Rule four: Leave nothing undefined
When in a trouble about how to describe a certain function, there is a
simple, but bad, way out. That is to leave the function undefined. This
is bad, because implementors may choose to implement this function in
different ways, which can cause interoperability problems.
10. Rule five: Cause no harm
Refractory implementers may follow a standards specification to the
letter, but interpret it in ways which means that other agents
receiving the information may be confused or even crash. Thus, the
"cause no harm" rule says that you should not produce information which
has a potential of causing harm to a well-working internet. Examples of
this is to send unreasonably long data, which might cause so-called
"buffer overflows" which are known to be a major cause of security
problems in standards. An example is lines in e-mail, where the
[RFC2822] standard says:
Each line of characters MUST be no more than 998 characters,
and SHOULD be no more than 78 characters, excluding the CRLF.
Note that it is possible to send e-mail with more than 78 characters
per line, using either the MIME [RFC2046] standard and the BASE64 or
the Quoted-Printable encodings, or by using the format=flowed format
[RFC 2646].
11. Rule six: Keep it simple, stupid
There is a tendency for standards developers to make standards overly
complex. One reason for this is that a simple way out of disagreements
and reaching consensus is to include everything anyone wants in a
standard.
There is also an obvious risk that people from large software vendors
may prefer a complex standard, since this reduces the number of
competitors who have the capacity to implement the full standard.
IETF is not immune from this tendency. For example the HTTP standard
does probably provide unnecessarily too many ways of indicating that a
page should not be automatically saved in any way in the receiving
computer. I read, in a bulletin board, the following recommendation of
how to use HTTP to reduce the risk that a page is saved:
Expires: Mon, 26 Jul 1997 05:00:00 GMT ; Date in the past
Last-Modified: 4 June 2005 15:20:00 GMT ; Current date
Cache-Control: no-store, nocache, must-revalidate, max-age=0 ; HTTP/1.1
Cache-Control: no-store, no-cache, must-revalidate, max-age=0 ; HTTP/1.1
Pragma: no-cache ; HTTP/1.0
The reason this recommendation is given, is that some HTTP
implementations may not support all the methods, so therefore it is
safest to use all of them. This is a typical example of the risk that
complex standards cause different implementations to support different
subsets of the standard, which can cause incompatibility problems.
Note: There are certainly reasons why the HTTP authors have chosen to
provide so many ways of specifying this.
In spite of what is said above, IETF standards are usually much simpler
than, for example, the competing OSI standards developed by ISO and
ITU. And this is probably a major reason why IETF standards usually
have been more successful than competing ISO and ITU standards.
The rule that any feature not implemented in two co-working independent
standards, before an IETF standard is progressed from "Proposed
standard" to "Draft standard" has certainly contributed to making the
standards simpler.
The advantage with simpler standards are that they are usually less
easy to misunderstand by implementers. Also, with a complex standard,
the risk is larger that two implementers will implement non-compatible
subsets of the standard. For ISO/ITU standards, a special set of
standards called "functional standards" specify good subsets to the
ISO/ITU standards. Such functional standards are not as much needed for
IETF standards, since they are simpler from the beginning.
12. Rule seven: No voting, rough consensus
IETF developers often claim that IETF has a rule of "no voting". Anyone
who has participated in IETF meetings know that this is not true, some
kinds of voting does happen now and then. Voting, however, is not part
of the formal decision process in IETF, as it is in ISO and ITU.
One reason for this is that all are not equal in standards work.
Opinions of people known to be reasonable and knowledgeable are more
important than those of other people participating in the standards
work.
Another reason is that there is no voting procedure which will always
produce the optimal result. This problem has been extensively discussed
in the political science sub-area with the name "social choice theory".
A simple example to illustrate the issue. Suppose we are to choose
between three solutions.
+----------+----------+----------+----------+----------+----------+
| |very good | good | OK | bad | very bad |
+----------+----------+----------+----------+----------+----------+
|Solution A| 40 % | | | | 60 % |
+----------+----------+----------+----------+----------+----------+
|Solution B| 35 % | | | | 65 % |
+----------+----------+----------+----------+----------+----------+
|Solution C| 25 % | 75 % | | | |
+----------+----------+----------+----------+----------+----------+
It is not obvious, in the above situation, what is the best choice to
make in a standards making organisation. Probably solution C is best,
since everyone thinks it is good, even though it has the least number
of people who thinks it is very good.
I have in fact implemented a voting program with the special goal of
being suitable for IETF. In this program, people are given a list of
alternatives, and can rate each alternative on a scale from very good
to very bad. The results are not presented as numbers. Instead, the
results are presented by listing the names of the people who voted in
each box in the table as shown in the example above.
"Rough consensus" means that all or most reasonable experts agree, even
though some may dissent.
13. Rule eight: Plain ASCII text is enough
This may be the least important of the golden rules, but all IETF
standards are written in plain ASCII, with no other formatting than
CRLF between the lines and FF between the pages. They are also written
to be printed with a monospace font (a font where all characters have
equal widths) otherwise some tables or diagrams (like in this memo) may
not display well.
This rule is not only for standards text. Also discussion about
standards in IETF working groups adhere to this rule. MIME, HTML and
other newfangled ideas are looked down upon in IETF mailing lists. And
best is if lines are no longer than 72 characters.
14. Contradiction between the rules
There may, of course, be contradictons between the rules. Note however
that Rule two and Rule three are rules about relaying information,
while Rule one and Rule five are rules about producing information. The
rest of the rules mainly concern how standards are written, rather than
how they are implemented.
A well-known example of a contradiction is if you have to combine
snipetts of HTML produced by different agents. An error in one of the
snippets, for example <font color=white> might then damage HTML coming
after the illegally formatted snippets. The best way around this
problem is to surround the included HTML with <DIV> and </DIV>, or,
even safer, put it into a separate FRAME or IFRAME (although FRAMEs and
IFRAMEs may have other drawbacks). Some software also adds finishing
HTML for unfinished elements, for example adds </P>, </FONT>, etc. at
the end of the incorrectly formatted HTML snippet. These additions may
be wrong, but the risk is certainly less than by adding start HTML
elements to try to correct incorrect snippets.
15. Other documents
There are many RFCs describing IETF workings and procedure. Some of
them are certainly more authorative than this memo. Examples of other
such RFCs are (in publication order):
RFC2418: IETF Working Group Guidelines and Procedures, by S. Bradner,
September 1998.
RFC3160: The Tao of IETF - A Novice's Guide to the Internet
Engineering, by S. Harris, August 2001.
RFC3184: IETF Guidelines for Conduct, by S. Harris. October 2001.
RFC3233: Defining the IETF, by P. Hoffman, S. Bradner. February 2002.
RFC3844: IETF Problem Resolution Process. E. Davies, edited by J.
Hofmann, August 2004.
RFC3929: Alternative Decision Making Processes for Consensus-Blocked
Decisions in the IETF, by T. Hardie. October 2004.
RFC3935: A Mission Statement for the IETF, by H. Alvestrand, October
2004.
16. References
For more references, see chapter 14 above.
[RFC1123]: Requirements for Internet Hosts -- Application and Support,
By Robert Braden, October 1989.
[RFC2046]: Multipurpose Internet Mail Extensions(MIME) Part Two: Media
Type, by N Freed and N Borenstein, November 1996.
[RFC2476]: Message swubmission, by R. Gellens and J. Klensin, December
1998.
[RFC2557]: MIME Encapsulation of Aggregate Documents, such as HTML
(MHTML), by J. Palme, A. Hoffman and N. Shellness, March 1999.
[RFC2646]: The Text/Plain Format and DelSp Parameter, by R. Gellens,
February 2004.
[RFC2822]: Internet Message Format, by Pete Resnick, April 2001.
17. Acknowledgements
Funding for the RFC Editor function is currently provided by the
Internet Society.
18. Author's address
Jacob Palme <jpalme@dsv.su.se>
Skeppargatan 73
11530 Stockholm
Phone: +46-8-16 16 67
19. Intellectual Property Statement
The IETF takes no position regarding the validity or scope of any
intellectual property or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; neither does it represent that it
has made any effort to identify any such rights. Information on the
IETF's procedures with respect to rights in standards-track and
standards-related documentation can be found in BCP-11. Copies of
claims of rights made available for publication and any assurances of
licenses to be made available, or the result of an attempt made to
obtain a general license or permission for the use of such
proprietary rights by implementers or users of this specification can
be obtained from the IETF Secretariat.
The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights which may cover technology that may be required to practice
this standard. Please address the information to the IETF Executive
Director.
20. Full Copyright Statement
Copyright (C) The Internet Society (2003). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assignees.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
This Internet-Draft will expire on 6 December 2005.