]> Cloudflare

lucas@lucaspardue.com

Apple

tpauly@apple.com

Armidale Consulting

dave.thaler.ietf@gmail.com

next generation unicorn sparkling distributed ledger Active use and maintenance of network protocols is an important way to ensure that protocols remain interoperable and extensible over time. Techniques such as intentionally exercising extension points with non-meaningful values (referred to as "grease") or adding variability to how protocol elements are used help generate this active use. Grease and variability are used across various protocols developed by the IETF. This document discusses considerations when designing and deploying grease and variability mechanisms, and provides advice for making them as effective as possible. About This Document The latest revision of this draft can be found at . Status information for this document may be found at . Source for this draft and an issue tracker can be found at .

Introduction discusses "active use" as a category of techniques that protocol designers and implementers employ to ensure that protocol extension mechanisms are exercised and can be used in the future. This ability to change (to handle protocol updates and extensions) is an important factor in the success of protocol deployment, as discussed in . Active use of protocol features and extensions often requires intentional efforts beyond what would organically occur in deployments. Some extension points do not frequently see new values being used, but are still important to be usable in the future. Some patterns of protocol usage might be relatively static without specific efforts to ensure that they can change in the future. One key techique for intentional use is "grease", or "greasing". Greasing was initially designed for TLS and was later adopted by other protocols such as QUIC . In these protocols, extension codepoints are reserved only for greasing and must be ignored upon receipt. Greasing is suitable for many protocols but not all; discusses the applicability and limitations of greasing. While it is becoming more common, designing and applying grease is not necessarily trivial. There are best practices, and some common pitfalls to avoid, that have been developed by the protocols using grease thus far. takes these learnings and provides considerations for new protocol design and deployment. Separate from greasing using reserved values, protocol deployments can intentionally add variability to ensure that network endpoints and middleboxes do not end up ossifying certain patterns. For example, an HTTP deployment focused on downloads might want to add support for uploads. Changing use of the application and transport protocol features can affect the deployment's network traffic profile. If expectations have been formed around historical patterns of use, i.e., ossification, introducing change might lead to deployment problems. presents considerations about how intentionally increasing the variability of protocols can mitigate some of these concerns. Protocol extensions can provide longevity in the face of changing needs or environment. However, a replacment protocol might be preferred when extensions are not adequate or feasible. A protocol replacement could aggregate common extensions and possibly make them mandatory, effectively defining a new baseline that can simplify deployment and interoperability. A replacement protocol version may or may not be compatible with other versions. A protocol may or may not have a mechanism for version selection or agility. presents considerations about designing for and/or implementing version negotiation and migration.

Considerations for Greasing Protocols Greasing can take many forms, depending on the protocol and the nature of its extension points. The common pattern across forms of greasing is that values are generated that have no useful meaning to the protocol and are meant to be ignored upon receipt. Such values used for the purpose of greasing are referred to as "grease values" within this document. More background to this approach is given in . This section provides some practical considerations for how to define and use greasing, and avoid possible pitfalls.

Define and Register Grease Value Ranges Many protocols that use greasing have a limited set of possible values or codepoints that can be used in a particular extension point. A common approach is to reserve a subset of the registrable space for greasing. The following are some examples of protocols that have reserved codepoints for grease values:

• TLS ()
• QUIC ()
• HTTP/3 ()
• Privacy Pass ()

The specifics of how to reserve values depends on the nature of the available space. For protocols with large possible spaces, it is useful to have a large set of grease values to increase the chance that receiver greasing requirements are exercised. The specific size and distribution of the grease range needs to accommodate the protocol constraints. For instance, protocols that use 8-bit fields may find it too costly to dedicate many grease values, while 32-bit or 64-bit fields are likely to have no such limitations. It is recommended to use an algorithm to reserve large sets of values. For example, uses an algorithm of 31 * N + 27 to allocate grease values for transport parameters. One possible problem with some algorithms is that they will spread out values over the space, and impact the ability to use or reserve contiguous blocks of non-grease values. It is common for protocol extension designers to want to reserve contiguous blocks of codepoints in order to aid iteration and experimentation. Reserved grease values can end up being in spaces that would otherwise be used for such contiguous blocks. Codepoints being used for new reservations, or experimentation, need to be careful to not unintentionally use grease values. Doing so could lead to interoperability failures.

Recommendations for IANA Considerations IANA registries that contain reserved grease values need to indicate that the values are reserved in order to prevent them from being allocated for other uses. The specifics of how to represent the reservations is up to the documents that define the registries. Some registries list out the reserved grease values individually, marked as "Reserved". For example, the TLS registry uses this approach (https://www.iana.org/assignments/tls-parameters/tls-parameters.xhtml). If an algorithm or pattern is used to define grease values, it is recommended when possible to instead only define a single entry for the entire grease value set. This entry should include the pattern or algorithm. This approach is used by the QUIC registry (https://www.iana.org/assignments/quic/quic.xhtml) and the HTTP/3 registry (https://www.iana.org/assignments/http3-parameters/http3-parameters.xhtml#http3-parameters-frame-types) Grease values must not be used or registered for any other purpose. This is in contrast to other potentially "reserved" values that might be reused or claimed for a new purpose in the future. To avoid confusion, it is recommended to label the reservation with a clear identifier, such as "reserved for greasing".

Use Unpredictable Grease Values In order to gain the benefits of active use and avoid ossification, grease values used in actual packets need to be sent in ways that won't become a predictable pattern that receiver and middlebox implementations and deployments ossify around. Implementations that generate grease values should pick unpredictable entries from the set of reserved grease values. It is most important that values be unpredictable across the set of all protocol participants for particular deployments. This can be achieved in multiple ways: for example, an individual sender might pick random values from the grease value space on each interaction; alternatively, a single sender could pick a specific grease value to use, while other senders pick other values. In order to support picking unpredictable values, the set of reserved values should be large, when possible. See for a discussion of how to allocate grease values.

Use Grease Values Unpredictably In addition to selecting unpredictable values, the inclusion of grease itself can be made unpredictable. Implementations can vary their behavior by including no grease values, one grease value, or multiple grease values for a given protocol extension point. How consistently and frequently to use grease values is a choice that implementations and deployments need to consider and weigh against several factors. Deployments of greasing should consider how they expect errors exposed by using grease values to be noticed and measured. If grease values are sent too infrequently, so that errors due to sending grease values blend in with the noise of other errors, it is likely that no one will notice failures, thus defeating the purpose of greasing. When grease values are sent more frequently, they will be noticed more. However, if grease values are sent too consistently, receiver implementations might end up special-casing grease values. The patterns for sending grease values can be made more effective by coordinating between devices sending the values. One example of coordination is having a "flag day" where implementations start sending grease values broadly, and measure to see where errors occur. See for more considerations concerning the use of a flag day.

Don't Handle Grease Values as a Special Case Implementations that read and process grease values must ignore the values. "Ignoring" a value upon receipt can have multiple dimensions, however. Simply not performing any protocol action based on the grease value isn't enough to ensure that the protocol will remain extensible. The ignoring must be handled as a general case of "unknown" or "unhandled" values, not as a special case for ignored grease values. This means that grease values can only meaningfully be used for protocol elements where all unknown values are ignored by default. (Protocols may have ways to indicate that some specific values are "mandatory" or "critical" in order to have the protocol interaction succeed, but these must not be marked for any grease values.) One pitfall that implementations may encounter when building logic to handle the receipt of grease values is related to cases where some recognized non-grease values need to be handled as errors. Consider the following abstract example:

1. A protocol element has values 1, 2, 3, and 4 defined and registered for use. The values 13 and 42 are reserved as grease values.
2. In a specific scenario, only the known values 1 and 2 are valid; 3 and 4 are considered errors.
3. An implementation might naively choose to check for the value being 1 or 2, handle those cases, and send an error otherwise.
4. When grease values are used, the previous logic will flag an error for the grease value. If this is detected, implementations might choose to work around this by updating the logic to check for the value being 1 or 2, then check for grease values to ignore, and then send an error otherwise. This logic is also incorrect since it doesn't allow for new extensibility.

The correct logic for the above scenario would be to check for the value being 1 or 2, then explicitly check for the value being 3 or 4 (to handle the error), with a catch-all for ignored values. In pseudo-code, the correct logic would work like this, where the grease values would fall into the final else case as ignored values. Implementations need to take care when implementing such logic. Protocol specification designers should emphasize that grease values must not be special-cased. It is also recommended to provide example logic or pseudocode in specifications, similar to the example above, as guidance to implementers on how to correctly process protocol elements like these. Documents can also provide test vectors, when applicable, that include grease values to ensure they are processed correctly. One limitation of greasing is that it only exercises grease values. It does not check whether other values that are reserved for future use are correctly treated as ignored or errors. As such, ossification remains a possibility for non-grease values. The goal of greasing is to increase the chances of correct implementation, and there by reduce (but not eliminate) the possibility of ossification.

Deployment Considerations and Incentives for Greasing Greasing can be used as a tool to improve the active use of existing protocol elements (which weren't necessarily designed with greasing to begin with, or weren't deployed with greasing); or as part of new protocol design and deployments. When greasing isn't used from the beginning of protocol deployment, starting to use greasing comes with the risk of triggering failures or anomalies. These failures might be innocuous, but they also might be very impactful and visible to users. This risk creates a disincentive to deploy greasing in existing systems, since generally the change that triggers failures is often blamed for the failure. The risk is highest when adding greasing to a particular protocol flow that doesn't require any change of behavior or adoption to hit the greasing behavior. For example, if a service migrates to use a new web server implementation that enables greasing, while the previous server didn't, some new failures may be hit if clients react poorly to greasing. Some approaches to avoid failures due to greasing include:

• Designing, implementing, and using greasing very early on in protocol development and deployment. This avoids the aforementioned risk of adding greasing late in a deployment.
• Enabling greasing along with other major protocol feature changes or deployment changes. For example, when upgrading to a new protocol version that requires implementation updates on multiple systems, greasing can be added for the new version specifically. This approach works well for situations where the protocol participants are known and already need to cooperate (such as within an encrypted protocol between two endpoints). This approach applies less well for situations where non-cooperating entities (such as middleboxes) are the source of ossification.
• Using heuristics to disable greasing when errors are encountered. For example, if a client-initiated protocol operation fails multiple times when grease values are used, it can be retried without any grease values. Alternatively, if a server recognizes a property of a client that always fails when greasing is used, it could choose to disable greasing when that client is detected. This reduces the effectiveness of grease values in removing existing ossification, but can still have benefits for flagging issues in new implementations when they receive grease values.

Discouraging Ossification Greasing alone can help ensure that middleboxes tolerate unrecognized extensions rather than crashing, but one common middlebox behavior is to simply drop all unrecognized extensions. Techniques such as disabling greasing when errors are encountered can enable communication but may not provide enough incentive to discourage ossification. One approach to preserving the incentive to grease without being blamed for doing so, while still potentially discouraging ossification, is to make any such ossification visible. For example, a sender can log and report ossification issues on a given path when falling back to non-grease values. Similarly, a receiver can log and report communication that does not include grease values. Such reports might be used in social feedback (e.g., public product or service reviews), or in contractual business relationship (e.g., SLA) discussion, in order to provide an incentive to implementers or deployers of ossified implementations.

Considerations for Increasing Protocol Variability Greasing can maintain protocol extensibility by falsifying active use of its extension points (see ). However, greasing alone does not ensure positive use of extension mechanisms. A protocol may define a wide-ranging extension capability that remains unused in the absence of real use cases. This can lead to ossification that does not expect extensions, leading to interoperability problems later on. Long-term maintenance and interoperability can be ensured by exercising extension points positively. To some extent this can be thought of as protocol fuzzing. This might be difficult to exercise because varying the protocol elements might change the outcome of interactions, leading to real errors. However, some protocols allow elements to be be safely changed, as shown in the following example.

Example: QUIC frames QUIC packets contain frames. Receivers might build expectations on the longitudinal aspects of packets or frames - size, ordering, frequency, etc. A sender can quite often manipulate these parameters and stay compliant to the requirements of the QUIC protocol. A QUIC stream is an ordered reliable byte stream that is serialized as a sequence of STREAM frames with a length and offset. Receivers are expected to reassemble frames, which could arrive in any order, into an ordered reliable byte stream that is readable by applications. A form of positive testing is for a sender to unpredictably order the STREAM frames that it transmits. For example, the sender can vary the sequence order of offset values. This allows exercising the QUIC reassembly features of the receiver with the expectation that no failure would occur. However, doing this may introduce delay or stream head-of-line blocking that affects the performance aspects of a transmission, which may not be acceptable for a given use case. As such, positive testing might be most appropriate to use in a subset of connections, or phases within a connection.

Considerations for Protocol Versions discusses considerations around planning for transitioning from an existing protocol, or protocol version, to a new one. There are also intrinsic and well-documented issues related to testing version negotiation of protocols; see and Sections and of . This section will be expanded with advice for protocol designers and implementers about how to approach these problems.

Security Considerations The considerations in , , , and all apply to the topics discussed in this document. The use of protocol features, extensions, and versions can already allow fingerprinting . Any techniques that change parameters in any way, including but not limited to those discussed in this document, can affect fingerprinting. A deeper analysis of this topic has been deemed out of scope.

IANA Considerations This document has no IANA actions itself. Guidance on how other documents can effectively instruct IANA about protocol greasing is provided in

Informative References The QUIC transport protocol has several features that are desirable in a transport for HTTP, such as stream multiplexing, per-stream flow control, and low-latency connection establishment. This document describes a mapping of HTTP semantics over QUIC. This document also identifies HTTP/2 features that are subsumed by QUIC and describes how HTTP/2 extensions can be ported to HTTP/3. The ability to change protocols depends on exercising the extension and version-negotiation mechanisms that support change. This document explores how regular use of new protocol features can ensure that it remains possible to deploy changes to a protocol. Examples are given where lack of use caused changes to be more difficult or costly. The Internet community has specified a large number of protocols to date, and these protocols have achieved varying degrees of success. Based on case studies, this document attempts to ascertain factors that contribute to or hinder a protocol's success. It is hoped that these observations can serve as guidance for future protocol work. This memo provides information for the Internet community. This document describes GREASE (Generate Random Extensions And Sustain Extensibility), a mechanism to prevent extensibility failures in the TLS ecosystem. It reserves a set of TLS protocol values that may be advertised to ensure peers correctly handle unknown values. This document defines the core of the QUIC transport protocol. QUIC provides applications with flow-controlled streams for structured communication, low-latency connection establishment, and network path migration. QUIC includes security measures that ensure confidentiality, integrity, and availability in a range of deployment circumstances. Accompanying documents describe the integration of TLS for key negotiation, loss detection, and an exemplary congestion control algorithm. This document defines an HTTP authentication scheme for Privacy Pass, a privacy-preserving authentication mechanism used for authorization. The authentication scheme specified in this document can be used by Clients to redeem Privacy Pass tokens with an Origin. It can also be used by Origins to challenge Clients to present Privacy Pass tokens. Over the many years since the introduction of the Internet Protocol, we have seen a number of transitions throughout the protocol stack, such as deploying a new protocol, or updating or replacing an existing protocol. Many protocols and technologies were not designed to enable smooth transition to alternatives or to easily deploy extensions; thus, some transitions, such as the introduction of IPv6, have been difficult. This document attempts to summarize some basic principles to enable future transitions, and it also summarizes what makes for a good transition plan. This document discusses architectural issues related to the extensibility of Internet protocols, with a focus on design considerations. It is intended to assist designers of both base protocols and extensions. Case studies are included. A companion document, RFC 4775 (BCP 125), discusses procedures relating to the extensibility of IETF protocols. This document is not an Internet Standards Track specification; it is published for informational purposes. The main goal of the networking standards process is to enable the long-term interoperability of protocols. This document describes active protocol maintenance, a means to accomplish that goal. By evolving specifications and implementations, it is possible to reduce ambiguity over time and create a healthy ecosystem. The robustness principle, often phrased as "be conservative in what you send, and liberal in what you accept", has long guided the design and implementation of Internet protocols. However, it has been interpreted in a variety of ways. While some interpretations help ensure the health of the Internet, others can negatively affect interoperability over time. When a protocol is actively maintained, protocol designers and implementers can avoid these pitfalls. This document explains why the IAB believes that, when there is a conflict between the interests of end users of the Internet and other parties, IETF decisions should favor end users. It also explores how the IETF can more effectively achieve this. This document offers guidance for developing privacy considerations for inclusion in protocol specifications. It aims to make designers, implementers, and users of Internet protocols aware of privacy-related design choices. It suggests that whether any individual RFC warrants a specific privacy considerations section will depend on the document's content.

Acknowledgments This work is a summary of the topics discussed during EDM meetings. The contributors at those meetings are thanked.