Network Working Group M. Thomson
Internet-Draft Mozilla
Intended status: Informational October 27, 2017
Expires: April 30, 2018
The Harmful Consequences of the Robustness Principle
draft-thomson-postel-was-wrong-02
Abstract
Jon Postel's famous statement in RFC 1122 of "Be liberal in what you
accept, and conservative in what you send" is a principle that has
long guided the design and implementation of Internet protocols. The
posture this statement advocates promotes interoperability, but can
produce negative effects in the protocol ecosystem in the long term.
Those effects can be avoided by properly maintaining protocols.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
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."
This Internet-Draft will expire on April 30, 2018.
Copyright Notice
Copyright (c) 2017 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
Thomson Expires April 30, 2018 [Page 1]
Internet-Draft Elephants Out, Donkeys In October 2017
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Fallibility of Specifications . . . . . . . . . . . . . . . . 3
3. Protocol Decay . . . . . . . . . . . . . . . . . . . . . . . 4
4. Ecosystem Effects . . . . . . . . . . . . . . . . . . . . . . 5
5. An Alternative Conclusion . . . . . . . . . . . . . . . . . . 6
6. The Role of Feedback . . . . . . . . . . . . . . . . . . . . 7
6.1. Fault Reporting is Valuable . . . . . . . . . . . . . . . 7
6.2. The Role of Strict Error Handling . . . . . . . . . . . . 7
7. Implementations Are Ultimately Responsible . . . . . . . . . 8
8. Security Considerations . . . . . . . . . . . . . . . . . . . 9
9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9
10. Informative References . . . . . . . . . . . . . . . . . . . 9
Appendix A. Acknowledgments . . . . . . . . . . . . . . . . . . 10
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 10
1. Introduction
Of the great many contributions Jon Postel made to the Internet, his
remarkable technical achievements are often ignored in favor of the
design and implementation philosophy of what is known as the
robustness principle:
Be strict when sending and tolerant when receiving.
Implementations must follow specifications precisely when sending
to the network, and tolerate faulty input from the network. When
in doubt, discard faulty input silently, without returning an
error message unless this is required by the specification.
This being the version of the text that appears in IAB RFC 1958
[PRINCIPLES].
In comparison, his contributions to the underpinnings of the
Internet, which are in many respects more significant, enjoy less
conscious recognition. Postel's robustness principle has been hugely
influential in shaping the Internet and the systems that use Internet
protocols. Many consider this principle to be instrumental in the
success of the Internet as well as the design of interoperable
protocols in general.
Over time, considerable changes have occurred in both the scale of
the Internet and the level of skill and experience available to
protocol and software designers. Much of that experience is with
Thomson Expires April 30, 2018 [Page 2]
Internet-Draft Elephants Out, Donkeys In October 2017
protocols that were designed, informed by Postel's maxim, in the
early phases of the Internet.
That experience shows that there are negative long-term consequences
to interoperability if an implementation applies Postel's advice.
Correcting the problems caused by divergent behavior in
implementations can be difficult.
This document describes the negative consequences of the application
of Postel's principle to the modern Internet. It recommends against
following Postel's advice, instead recommending that protocols
receive continuing maintenance after their initial design and
deployment. Active maintenance of protocols reduces or eliminates
the opportunities to apply Postel's guidance.
There is good evidence to suggest that protocols are routinely
maintained beyond their inception. This document serves primarily as
a record of the shortcomings of the robustness principle.
2. Fallibility of Specifications
What is often missed in discussions of the robustness principle is
the context in which it appears. The earliest form of the principle
in the RFC series (in RFC 760 [RFC0760]) is preceded by a sentence
that reveals the motivation for the principle:
While the goal of this specification is to be explicit about the
protocol there is the possibility of differing interpretations.
In general, an implementation should be conservative in its
sending behavior, and liberal in its receiving behavior.
This motivating statement is a frank admission of fallibility and
remarkable for it. Here Postel recognizes the possibility that the
specification could be imperfect. This is an important statement,
but inexplicably absent from the later versions in [HOSTS] and
[PRINCIPLES].
Indeed, an imperfect specification is natural, largely because it was
- and remains thus - more important to proceed to implementation and
deployment than it is to perfect the protocol specification. A
protocol, like software, benefits greatly from experience in
deployment, and a deployed protocol is immeasurably more useful than
a perfect protocol.
As shown [SUCCESS] demonstrates, success or failure of a protocol
depends far more on factors like usefulness than on on technical
excellence. Postel's timely publication of protocol specifications,
Thomson Expires April 30, 2018 [Page 3]
Internet-Draft Elephants Out, Donkeys In October 2017
even with the potential for flaws, likely had a significant effect in
the eventual success of the Internet.
The problem is therefore not with the premise, but with its
conclusion: the robustness principle itself.
3. Protocol Decay
Divergent implementations of a specification emerge over time. When
variations occur in the interpretation or expression of semantic
components, implementations cease to be perfectly interoperable.
Implementation bugs are often identified as the cause of variation,
though it is often a combination of factors. Application of a
protocol to new and unanticipated uses, and ambiguities or errors in
the specification are often confounding factors. Situations where
two peers disagree on interpretation should be expected over the
lifetime of a protocol.
Even with the best intentions, the pressure to interoperate can be
significant. No implementation can hope to avoid having to trade
correctness for interoperability indefinitely.
An implementation that reacts to variations in the manner advised by
Postel sets up a feedback cycle:
o Over time, implementations progressively add new code to constrain
how data is transmitted, or to permit variations in what is
received.
o Errors in implementations, or confusion about semantics can
thereby be masked.
o These errors can become entrenched, forcing other implementations
to be tolerant of those errors.
In this way an flaw can become entrenched as a de facto standard.
Any implementation of the protocol is required to replicate the
aberrant behavior, or it is not interoperable. This is both a
consequence of applying Postel's advice, and a product of a natural
reluctance to avoid fatal error conditions. Ensuring
interoperability in this environment is often colloquially referred
to as aiming to be "bug for bug compatible".
For example, in TLS [RFC5246] messages use an extension format with a
tag-length-value format. TLS extensions can be added to handshake
messages in any order. However, some server implementations
terminate connections if they encounter a TLS ClientHello message
Thomson Expires April 30, 2018 [Page 4]
Internet-Draft Elephants Out, Donkeys In October 2017
that ends with an empty extension. Thus, client implementations are
required to be aware of this incompatibility and ensure that a
ClientHello message ends in a non-empty extension.
While TLS highlights the potential for implementation error to cause
problems, the original JSON specification [RFC4627] demonstrates the
effect of specification shortcomings. RFC 4627 omitted critical
details on a range of key details including Unicode handling,
ordering and duplication of object members, and number encoding.
Consequently, a range of interpretations were used by
implementations. An updated JSON specification [RFC7159] was unable
to correct these errors, instead concentrating on defining the
interoperable subset of JSON. I-JSON [RFC7493] defines a new format
that is substantially similar to JSON while prohibiting the
problematic variations. However, that prohibition means that I-JSON
is not fully interoperable: an I-JSON implementation will fail to
accept some valid JSON texts. Consequently, I-JSON is not widely
implemented in parsers. Many JSON parsers instead implement the more
precise algorithm specified in [ECMA262].
It is debatable as to whether decay can be completely avoided, but
the robustness principle encourages a reaction that compounds
problems.
4. Ecosystem Effects
Once deviations become entrenched, it can be extremely difficult - if
not impossible - to rectify the situation.
For widely used protocols, the massive scale of the Internet makes
large-scale interoperability testing infeasible for all but a
privileged few. As the set of changes needed to maintain
interoperability grows in size, the cost of building a new
implementation increases. This is particularly relevant as the set
of tweaks necessary for interoperability can be difficult to learn.
Consequently, new implementations can be restricted to niche uses,
where the problems arising from interoperability issues can be more
closely managed. Restricting new implementations to narrow contexts
also risks causing forks in the protocol. If implementations cannot
interoperate, the chances of the addition of further incompatible
changes is significant.
This has a negative impact on the ecosystem of a protocol. New
implementations are important in ensuring the continued viability of
a protocol. New protocol implementations are also more likely to be
developed for new and diverse use cases and often are the origin of
features and capabilities that can be of benefit to existing users.
Thomson Expires April 30, 2018 [Page 5]
Internet-Draft Elephants Out, Donkeys In October 2017
The need to work around interoperability problems also reduce the
ability of established implementations to change. For instance, an
accumulation of mitigations for interoperability issues makes
implementations more difficult to maintain.
5. An Alternative Conclusion
The robustness principle is best suited to safeguarding against flaws
in a specification that is intended to remain unchanged for an
extended period of time. Indeed, in the face of divergent
interpretations of an immutable specification, the only hope for an
implementation to remain interoperable is to be tolerant of
differences in interpretation and occasional outright implementation
errors.
From this perspective, application of Postel's advice to the
implementation of a protocol specification that does not change is
logical, even necessary. But that suggests that the problem is with
the presumption of immutability of specifications.
Active maintenance of a protocol can ensure that specifications
remain accurate and that new implementations are possible. A
maintained protocol is not a static construct, it improves and
changes as it is used.
Protocol designers are strongly encouraged to continue to maintain
and evolve protocols beyond their initial inception and definition.
Maintenance is needed in response to the discovery of errors in
specification that might cause interoperability issues. Maintenance
is also critical for ensuring that the protocol is viable for
application to use cases that might not have been envisaged during
its original design. New use cases are an indicator that the
protocol could be successful [SUCCESS].
Maintenance does not demand the development of new versions of
protocols or protocol specifications. For instance, RFC 793 [TCP]
remains the canonical TCP reference, but that document alone is no
longer sufficient to implement the protocol. TCP is the subject of a
very large number of update and extension RFCs that together document
the deployed protocol.
Good extensibility [EXT] can make it easier to respond to new use
cases or changes in the environment in which the protocol is
deployed.
The process of maintenance ideally begins even before the
specification for a protocol is complete. Neglect can quickly
produce the negative consequences this document describes. Restoring
Thomson Expires April 30, 2018 [Page 6]
Internet-Draft Elephants Out, Donkeys In October 2017
the protocol to a state where it can be maintained involves first
discovering the properties of the protocol as it is deployed, rather
than the protocol as it was originally documented. This can be
difficult and time-consuming, particularly if the protocol has a
diverse set of implementations. Such a process was undertaken for
HTTP [HTTP] after a period of minimal maintenance. Restoring the
specification to currency took significant effort over more than 6
years.
6. The Role of Feedback
Protocol maintenance is only possible if there is sufficient
information about the deployment of the protocol. Feedback from
deployment is critical to effective protocol maintenance.
For a protocol specification, the primary and most effective form of
feedback comes from people who implement or deploy the protocol. An
active community of protocol implementers and users is the most
valuable source of feedback. It is this community that implements
and deploys changes, in addition to contributing to the ongoing
maintenance of protocol specifications.
6.1. Fault Reporting is Valuable
Protocol implementations should include automated error reporting
mechanisms.
Exposing faults through operations and management systems is highly
valuable, but it might be necessary to ensure that the information is
propagated further.
Building telemetry and error logging systems that report faults to
the developers of the implementation is superior in many respects.
However, this is only possible in deployments that are conducive to
the collection of this type of information. Giving consideration to
protection of the privacy of protocol participants is critical prior
to deploying any such system.
6.2. The Role of Strict Error Handling
Favoring strict error handling over attempting error recovery is an
effective technique for ensuring that faults receive attention.
A fatal errors provide excellent motivation to address problems.
However, generating fatal errors is only feasible if such errors are
sufficiently rare. Frequent problems can result in users to ignoring
them or finding workarounds, whereas rare events demand greater
attention.
Thomson Expires April 30, 2018 [Page 7]
Internet-Draft Elephants Out, Donkeys In October 2017
Fatal errors or crashes are also preferable if there is any risk that
the error might represent an implementation or security issue.
Exposing errors is particularly important for early implementations
of a protocol. Enabling stricter error handling helps validate the
design of both protocol and implementations. If strict checks are
retained when implementations are more widely deployed, those checks
better enable the detection and correction of errors in new
implementations.
This doesn't mean that protocol implementations need to treat all
inputs equally strictly. A protocol could be designed with the
explicit goal of accepting a wide range of inputs (see for example
[HTML]). As long as proper reactions to inputs are clearly and
unambiguously specified, these considerations don't apply.
7. Implementations Are Ultimately Responsible
Implementations and deployments are critical to the ongoing
maintenance of a protocol. Implementations have ample motivation to
prefer stability and interoperability over maintenance or
correctness. It is likely that - over time - an implementation will
accumulate allowances for errors of specification or implementation.
Without care, this can lead to suppression of feedback about
problems.
Even when an implementation chooses to attempt to adapt to an
abnormal condition, communicating that decision to the community is
valuable. At a minimum, it allows others to benefit from knowledge
of the problem. Discussion about problems might also lead to
alternative strategies or protocol enhancements that can avoid
further problems.
Over time, mitigations for interoperability issues could become
redundant. Occasional review of any special interoperability
measures might reveal opportunities for an implementation to switch
to stricter handling of exceptional conditions. Removing special
allowances in favor of stricter error reporting restores feedback
measures that new implementations can then benefit from.
Managing and deploying changes can be expensive. However, it is
widely recognized that maintenance is a critical part of the
deployment of computer systems for security reasons [IOTSU].
Managing updates for interoperability problems represents a small
additional cost in exchange for ensuring the ability to interoperate
with other users of the network.
Thomson Expires April 30, 2018 [Page 8]
Internet-Draft Elephants Out, Donkeys In October 2017
8. Security Considerations
Sloppy implementations, lax interpretations of specifications, and
uncoordinated extrapolation of requirements to cover gaps in
specification can result in security problems. Hiding the
consequences of protocol variations encourages the hiding of issues,
which can conceal bugs and make them difficult to discover.
Designers and implementers of security protocols generally understand
these concerns. However, general-purpose protocols are not exempt
from careful consideration of security issues. Furthermore, because
general-purpose protocols tend to deal with flaws or obsolescence in
a less urgent fashion than security protocols, there can be fewer
opportunities to correct problems in protocols that develop
interoperability problems.
9. IANA Considerations
This document has no IANA actions.
10. Informative References
[ECMA262] "ECMAScript(R) 2017 Language Specification", ECMA-262 8th
Edition, June 2017, <http://www.ecma-
international.org/publications/standards/Ecma-262.htm>.
[EXT] Carpenter, B., Aboba, B., Ed., and S. Cheshire, "Design
Considerations for Protocol Extensions", RFC 6709,
DOI 10.17487/RFC6709, September 2012,
<https://www.rfc-editor.org/info/rfc6709>.
[HOSTS] Braden, R., Ed., "Requirements for Internet Hosts -
Communication Layers", STD 3, RFC 1122,
DOI 10.17487/RFC1122, October 1989,
<https://www.rfc-editor.org/info/rfc1122>.
[HTML] "HTML", WHATWG Living Standard, October 2017,
<https://html.spec.whatwg.org/>.
[HTTP] Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer
Protocol (HTTP/1.1): Message Syntax and Routing",
RFC 7230, DOI 10.17487/RFC7230, June 2014,
<https://www.rfc-editor.org/info/rfc7230>.
[IOTSU] Tschofenig, H. and S. Farrell, "Report from the Internet
of Things Software Update (IoTSU) Workshop 2016",
RFC 8240, DOI 10.17487/RFC8240, September 2017,
<https://www.rfc-editor.org/info/rfc8240>.
Thomson Expires April 30, 2018 [Page 9]
Internet-Draft Elephants Out, Donkeys In October 2017
[PRINCIPLES]
Carpenter, B., Ed., "Architectural Principles of the
Internet", RFC 1958, DOI 10.17487/RFC1958, June 1996,
<https://www.rfc-editor.org/info/rfc1958>.
[RFC0760] Postel, J., "DoD standard Internet Protocol", RFC 760,
DOI 10.17487/RFC0760, January 1980,
<https://www.rfc-editor.org/info/rfc760>.
[RFC4627] Crockford, D., "The application/json Media Type for
JavaScript Object Notation (JSON)", RFC 4627,
DOI 10.17487/RFC4627, July 2006,
<https://www.rfc-editor.org/info/rfc4627>.
[RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security
(TLS) Protocol Version 1.2", RFC 5246,
DOI 10.17487/RFC5246, August 2008,
<https://www.rfc-editor.org/info/rfc5246>.
[RFC7159] Bray, T., Ed., "The JavaScript Object Notation (JSON) Data
Interchange Format", RFC 7159, DOI 10.17487/RFC7159, March
2014, <https://www.rfc-editor.org/info/rfc7159>.
[RFC7493] Bray, T., Ed., "The I-JSON Message Format", RFC 7493,
DOI 10.17487/RFC7493, March 2015,
<https://www.rfc-editor.org/info/rfc7493>.
[SUCCESS] Thaler, D. and B. Aboba, "What Makes for a Successful
Protocol?", RFC 5218, DOI 10.17487/RFC5218, July 2008,
<https://www.rfc-editor.org/info/rfc5218>.
[TCP] Postel, J., "Transmission Control Protocol", STD 7,
RFC 793, DOI 10.17487/RFC0793, September 1981,
<https://www.rfc-editor.org/info/rfc793>.
Appendix A. Acknowledgments
Constructive feedback on this document has been provided by a
surprising number of people including Mark Nottingham, Brian
Trammell, and Anne Van Kesteren. Please excuse any omission.
Author's Address
Martin Thomson
Mozilla
Email: martin.thomson@gmail.com
Thomson Expires April 30, 2018 [Page 10]