Internet-Draft | TAPS Interface | March 2020 |
Trammell, et al. | Expires 10 September 2020 | [Page] |
This document describes an abstract programming interface to the transport layer, following the Transport Services Architecture. It supports the asynchronous, atomic transmission of messages over transport protocols and network paths dynamically selected at runtime. It is intended to replace the traditional BSD sockets API as the lowest common denominator interface to the transport layer, in an environment where endpoints have multiple interfaces and potential transport protocols to select from.¶
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 10 September 2020.¶
Copyright (c) 2020 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 the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
This document specifies a modern abstract programming interface atop the high-level architecture for transport services defined in [I-D.ietf-taps-arch]. It supports the asynchronous, atomic transmission of messages over transport protocols and network paths dynamically selected at runtime. It is intended to replace the traditional BSD sockets API as the lowest common denominator interface to the transport layer, in an environment where endpoints have multiple interfaces and potential transport protocols to select from.¶
As applications adopt this interface, they will benefit from a wide set of transport features that can evolve over time, and ensure that the system providing the interface can optimize its behavior based on the application requirements and network conditions, without requiring changes to the applications. This flexibility enables faster deployment of new features and protocols. It can also support applications by offering racing and fallback mechanisms, which otherwise need to be implemented in each application separately.¶
It derives specific path and protocol selection properties and supported transport features from the analysis provided in [RFC8095], [I-D.ietf-taps-minset], and [I-D.ietf-taps-transport-security]. The design encourages implementations underneath the interface to dynamically choose a transport protocol depending on an application's choices rather than statically binding applications to a protocol at compile time. We note that transport system implementations SHOULD provide applications a way to override transport selection and instantiate a specific stack, e.g. to support servers wanting to listen to a specific protocol. This specific transport stack choice is discouraged for general use, as it comes at the cost of reduced portability.¶
This API is described in terms of Objects, which an application can interact with; Actions the application can perform on these Objects; Events, which an Object can send to an application asynchronously; and Parameters associated with these Actions and Events.¶
The following notations, which can be combined, are used in this document:¶
Object := Action()¶
[]Object := Action()¶
Object.Action()¶
Object -> Event<>¶
Action(param0, param1?, ...) / Event<param0, param1, ...>¶
Actions associated with no Object are Actions on the abstract interface itself; they are equivalent to Actions on a per-application global context.¶
How these abstract concepts map into concrete implementations of this API in a given language on a given platform is largely dependent on the features of the language and the platform. Actions could be implemented as functions or method calls, for instance, and Events could be implemented via event queues, handler functions or classes, communicating sequential processes, or other asynchronous calling conventions.¶
This specification treats Events and errors similarly. Errors, just as any other Events, may occur asynchronously in network applications. However, it is recommended that implementations of this interface also return errors immediately, according to the error handling idioms of the implementation platform, for errors that can be immediately detected, such as inconsistency in Transport Properties. Errors can provide an optional reason to give the application further details as to why the error occured.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The design of the interface specified in this document is based on a set of princples, themselves an elaboration on the architectural design principles defined in [I-D.ietf-taps-arch]. The interface defined in this document provides:¶
The Transport Services API is the basic common abstract application programming interface to the Transport Services Architecture defined in the TAPS Architecture [I-D.ietf-taps-arch].¶
An application primarily interacts with this API through two Objects: Preconnections and Connections. A Preconnection represents a set of properties and constraints on the selection and configuration of paths and protocols to establish a Connection with a remote Endpoint. A Connection represents a transport Protocol Stack on which data can be sent to and/or received from a remote Endpoint (i.e., depending on the kind of transport, connections can be bi-directional or unidirectional). Connections can be created from Preconnections in three ways: by initiating the Preconnection (i.e., actively opening, as in a client), through listening on the Preconnection (i.e., passively opening, as in a server), or rendezvousing on the Preconnection (i.e. peer to peer establishment).¶
Once a Connection is established, data can be sent and received on it in the form of Messages. The interface supports the preservation of message boundaries both via explicit Protocol Stack support, and via application support through a Message Framer which finds message boundaries in a stream. Messages are received asynchronously through event handlers registered by the application. Errors and other notifications also happen asynchronously on the Connection. It is not necessary for an application to handle all events; some events may have implementation-specific default handlers. The application should not assume that ignoring events (e.g. errors) is always safe.¶
Section 5, Section 6, Section 7, Section 8, and Section 11 describe the details of application interaction with Objects through Actions and Events in each phase of a Connection, following the phases (Pre-Establishment, Establishment, Data Transfer, and Termination) described in Section 4.1 of [I-D.ietf-taps-arch].¶
The following usage examples illustrate how an application might use a Transport Services Interface to:¶
The examples in this section presume that a transport protocol is available between the endpoints that provides Reliable Data Transfer, Preservation of data ordering, and Preservation of Message Boundaries. In this case, the application can choose to receive only complete messages.¶
If none of the available transport protocols provides Preservation of Message Boundaries, but there is a transport protocol that provides a reliable ordered byte stream, an application may receive this byte stream as partial Messages and transform it into application-layer Messages. Alternatively, an application may provide a Message Framer, which can transform a byte stream into a sequence of Messages (Section 9).¶
This is an example of how an application might listen for incoming Connections using the Transport Services Interface, receive a request, and send a response.¶
LocalSpecifier := NewLocalEndpoint() LocalSpecifier.WithInterface("any") LocalSpecifier.WithService("https") TransportProperties := NewTransportProperties() TransportProperties.Require(preserve-msg-boundaries) // Reliable Data Transfer and Preserve Order are Required by default SecurityParameters := NewSecurityParameters() SecurityParameters.AddIdentity(identity) SecurityParameters.AddPrivateKey(privateKey, publicKey) // Specifying a remote endpoint is optional when using Listen() Preconnection := NewPreconnection(LocalSpecifier, TransportProperties, SecurityParameters) Listener := Preconnection.Listen() Listener -> ConnectionReceived<Connection> // Only receive complete messages in a Conn.Received handler Connection.Receive() Connection -> Received<messageDataRequest, messageContext> //---- Receive event handler begin ---- Connection.Send(messageDataResponse) Connection.Close() // Stop listening for incoming Connections // (this example supports only one Connection) Listener.Stop() //---- Receive event handler end ----¶
This is an example of how an application might connect to a remote application using the Transport Services Interface, send a request, and receive a response.¶
RemoteSpecifier := NewRemoteEndpoint() RemoteSpecifier.WithHostname("example.com") RemoteSpecifier.WithService("https") TransportProperties := NewTransportProperties() TransportProperties.Require(preserve-msg-boundaries) // Reliable Data Transfer and Preserve Order are Required by default SecurityParameters := NewSecurityParameters() TrustCallback := NewCallback({ // Verify identity of the remote endpoint, return the result }) SecurityParameters.SetTrustVerificationCallback(TrustCallback) // Specifying a local endpoint is optional when using Initiate() Preconnection := NewPreconnection(RemoteSpecifier, TransportProperties, SecurityParameters) Connection := Preconnection.Initiate() Connection -> Ready<> //---- Ready event handler begin ---- Connection.Send(messageDataRequest) // Only receive complete messages Connection.Receive() //---- Ready event handler end ---- Connection -> Received<messageDataResponse, messageContext> // Close the Connection in a Receive event handler Connection.Close()¶
This is an example of how an application might establish a connection with a peer using Rendezvous(), send a Message, and receive a Message.¶
LocalSpecifier := NewLocalEndpoint() LocalSpecifier.WithPort(9876) RemoteSpecifier := NewRemoteEndpoint() RemoteSpecifier.WithHostname("example.com") RemoteSpecifier.WithPort(9877) TransportProperties := NewTransportProperties() TransportProperties.Require(preserve-msg-boundaries) // Reliable Data Transfer and Preserve Order are Required by default SecurityParameters := NewSecurityParameters() SecurityParameters.AddIdentity(identity) SecurityParameters.AddPrivateKey(privateKey, publicKey) TrustCallback := New Callback({ // Verify identity of the remote endpoint, return the result }) SecurityParameters.SetTrustVerificationCallback(trustCallback) // Both local and remote endpoint must be specified Preconnection := NewPreconnection(LocalSpecifier, RemoteSpecifier, TransportProperties, SecurityParameters) Preconnection.Rendezvous() Preconnection -> RendezvousDone<Connection> //---- Ready event handler begin ---- Connection.Send(messageDataRequest) // Only receive complete messages Connection.Receive() //---- Ready event handler end ---- Connection -> Received<messageDataResponse, messageContext> // Close the Connection in a Receive event handler Connection.Close()¶
Each application using the Transport Services Interface declares its preferences for how the transport service should operate using properties at each stage of the lifetime of a connection using Transport Properties, as defined in [I-D.ietf-taps-arch].¶
Transport Properties are divided into Selection, Connection, and Message Properties. During pre-establishment, Selection Properties (see Section 5.2) are used to specify which paths and protocol stacks can be used and are preferred by the application, and Connection Properties (see Section 10.1) can be used to influence decisions made during establishment and to fine-tune the eventually established connection. These Connection Properties can also be used later, to monitor and fine-tune established connections. The behavior of the selected protocol stack(s) when sending Messages is controlled by Message Properties (see Section 7.5).¶
All Transport Properties, regardless of the phase in which they are used, are organized within a single namespace. This enables setting them as defaults in earlier stages and querying them in later stages:¶
Note that configuring Connection Properties and Message Properties on Preconnections is preferred over setting them later. Early specification of Connection Properties allows their use as additional input to the selection process. Protocol Specific Properties, which enable configuration of specialized features of a specific protocol, see Section 3.2 of [I-D.ietf-taps-arch], are not used as an input to the selection process but only support configuration if the respective prototocol has been selected.¶
Transport Properties are referred to by property names. These names are lower-case strings whereby words are separated by hyphens. These names serve two purposes:¶
Transport Property Names are hierarchically organized in the form [<Namespace>.]<PropertyName>.¶
Namespaces for the keywords provided in the IANA protocol numbers registry (see https://www.iana.org/assignments/protocol-numbers/protocol-numbers.xhtml) are reserved for Protocol Specific Properties and MUST not be used for vendor or implementation specific properties.¶
Transport Properties can have one of a set of data types:¶
This document defines a language- and platform-independent interface to a Transport Services system. Given the wide variety of languages and language conventions used to write applications that use the transport layer to connect to other applications over the Internet, this independence makes this interface necessarily abstract.¶
There is no interoperability benefit in tightly defining how the interface is presented to application programmers across diverse platforms. However, maintaining the "shape" of the abstract interface across these platforms reduces the effort for programmers who learn the transport services interface to then apply their knowledge across multiple platforms.¶
We therefore make the following recommendations:¶
The Pre-Establishment phase allows applications to specify properties for the Connections they are about to make, or to query the API about potential Connections they could make.¶
A Preconnection Object represents a potential Connection. It has state that describes properties of a Connection that might exist in the future. This state comprises Local Endpoint and Remote Endpoint Objects that denote the endpoints of the potential Connection (see Section 5.1), the Selection Properties (see Section 5.2), any preconfigured Connection Properties (Section 10.1), and the security parameters (see Section 5.3):¶
Preconnection := NewPreconnection(LocalEndpoint?, RemoteEndpoint?, TransportProperties, SecurityParams)¶
The Local Endpoint MUST be specified if the Preconnection is used to Listen() for incoming Connections, but is OPTIONAL if it is used to Initiate() connections. The Remote Endpoint MUST be specified if the Preconnection is used to Initiate() Connections, but is OPTIONAL if it is used to Listen() for incoming Connections. The Local Endpoint and the Remote Endpoint MUST both be specified if a peer-to-peer Rendezvous is to occur based on the Preconnection.¶
Message Framers (see Section 9), if required, should be added to the Preconnection during pre-establishment.¶
The transport services API uses the Local Endpoint and Remote Endpoint Objects to refer to the endpoints of a transport connection. Actions on these Objects can be used to represent various different types of endpoint identifiers, such as IP addresses, DNS names, and interface names, as well as port numbers and service names.¶
Specify a Remote Endpoint using a hostname and service name:¶
RemoteSpecifier := NewRemoteEndpoint() RemoteSpecifier.WithHostname("example.com") RemoteSpecifier.WithService("https")¶
Specify a Remote Endpoint using an IPv6 address and remote port:¶
RemoteSpecifier := NewRemoteEndpoint() RemoteSpecifier.WithIPv6Address(2001:db8:4920:e29d:a420:7461:7073:0a) RemoteSpecifier.WithPort(443)¶
Specify a Remote Endpoint using an IPv4 address and remote port:¶
RemoteSpecifier := NewRemoteEndpoint() RemoteSpecifier.WithIPv4Address(192.0.2.21) RemoteSpecifier.WithPort(443)¶
Specify a Local Endpoint using a local interface name and local port:¶
LocalSpecifier := NewLocalEndpoint() LocalSpecifier.WithInterface("en0") LocalSpecifier.WithPort(443)¶
As an alternative to specifying an interface name for the Local Endpoint, an application can express more fine-grained preferences using the "Interface Instance or Type" Selection Property, see Section 5.2.10. However, if the application specifies Selection Properties which are inconsistent with the Local Endpoint, this will result in an error once the application attempts to open a Connection.¶
Specify a Local Endpoint using a STUN server:¶
LocalSpecifier := NewLocalEndpoint() LocalSpecifier.WithStunServer(address, port, credentials)¶
Specify a Local Endpoint using a Any-Source Multicast group to join on a named local interface:¶
LocalSpecifier := NewLocalEndpoint() LocalSpecifier.WithIPv4Address(233.252.0.0) LocalSpecifier.WithInterface("en0")¶
Source-Specific Multicast requires setting both a Local and Remote Endpoint:¶
LocalSpecifier := NewLocalEndpoint() LocalSpecifier.WithIPv4Address(232.1.1.1) LocalSpecifier.WithInterface("en0") RemoteSpecifier := NewRemoteEndpoint() RemoteSpecifier.WithIPv4Address(192.0.2.22)¶
Implementations may also support additional endpoint representations and provide a single NewEndpoint() call that takes different endpoint representations.¶
Multiple endpoint identifiers can be specified for each Local Endpoint and Remote Endpoint. For example, a Local Endpoint could be configured with two interface names, or a Remote Endpoint could be specified via both IPv4 and IPv6 addresses. These multiple identifiers refer to the same transport endpoint.¶
The transport services API resolves names internally, when the Initiate(), Listen(), or Rendezvous() method is called to establish a Connection. The API explicitly does not require the application to resolve names, though there is a tradeoff between early and late binding of addresses to names. Early binding allows the API implementation to reduce connection setup latency, at the cost of potentially limited scope for alternate path discovery during Connection establishment, as well as potential additional information leakage about application interest when used with a resolution method (such as DNS without TLS) which does not protect query confidentiality.¶
The Resolve() action on Preconnection can be used by the application to force early binding when required, for example with some Network Address Translator (NAT) traversal protocols (see Section 6.3).¶
Specifying a multicast group address on the Local Endpoint will indicate to the transport system that the resulting connection will be used to receive multicast messages. The Remote Endpoint can be used to filter by specific senders. This will restrict the application to establishing the Preconnection by calling Listen(). The accepted Connections are receive-only.¶
Similarly, specifying a multicast group address on the Remote Endpoint will indicate that the resulting connection will be used to send multicast messages.¶
A Preconnection Object holds properties reflecting the application's requirements and preferences for the transport. These include Selection Properties for selecting protocol stacks and paths, as well as Connection Properties for configuration of the detailed operation of the selected Protocol Stacks.¶
The protocol(s) and path(s) selected as candidates during establishment are determined and configured using these properties. Since there could be paths over which some transport protocols are unable to operate, or remote endpoints that support only specific network addresses or transports, transport protocol selection is necessarily tied to path selection. This may involve choosing between multiple local interfaces that are connected to different access networks.¶
Most Selection Properties are represented as preferences, which can have one of five preference levels:¶
Preference | Effect |
---|---|
Require | Select only protocols/paths providing the property, fail otherwise |
Prefer | Prefer protocols/paths providing the property, proceed otherwise |
Ignore | No preference |
Avoid | Prefer protocols/paths not providing the property, proceed otherwise |
Prohibit | Select only protocols/paths not providing the property, fail otherwise |
In addition, the pseudo-level Default
can be used to reset the property to the default
level used by the implementation. This level will never show up when queuing the value of
a preference - the effective preference must be returned instead.¶
The implementation MUST ensure an outcome that is consistent with application requirements as expressed using Require and Prohibit. While preferences expressed using Prefer and Avoid influence protocol and path selection as well, outcomes may vary given the same Selection Properties, as the available protocols and paths may vary across systems and contexts. However, implementations are RECOMMENDED to aim to provide a consistent outcome to an application, given the same Selection Properties.¶
Note that application preferences may conflict with each other. For example, if an application indicates a preference for a specific path by specifying an interface, but also a preference for a protocol, a situation might occur in which the preferred protocol is not available on the preferred path. In such cases, implementations SHOULD prioritize Selection Properties that select paths over those that select protocols. Therefore, the transport system SHOULD race the path first, ignoring the protocol preference if the protocol does not work on the path.¶
Selection and Connection Properties, as well as defaults for Message Properties, can be added to a Preconnection to configure the selection process and to further configure the eventually selected protocol stack(s). They are collected into a TransportProperties object to be passed into a Preconnection object:¶
TransportProperties := NewTransportProperties()¶
Individual properties are then added to the TransportProperties Object:¶
TransportProperties.Add(property, value)¶
Selection Properties of type "Preference" can be frequently used. Implementations MAY therefore provide additional convenience functions, see Appendix A.1 for examples. In addition, implementations MAY provide a mechanism to create TransportProperties objects that are preconfigured for common use cases as outlined in Appendix A.2.¶
For an existing Connection, the Transport Properties can be queried any time by using the following call on the Connection Object:¶
TransportProperties := Connection.GetTransportProperties()¶
A Connection gets its Transport Properties either by being explicitly configured via a Preconnection, by configuration after establishment, or by inheriting them from an antecedent via cloning; see Section 6.4 for more.¶
Section 10.1 provides a list of Connection Properties, while Selection
Properties are listed in the subsections below. Note that many properties are
only considered during establishment, and can not be changed after a Connection
is established; however, they can be queried. Querying a Selection Property
after establishment yields the value Require
for properties of the selected
protocol and path, Avoid for properties avoided during selection, and Ignore for
all other properties.¶
An implementation of this interface must provide sensible defaults for Selection Properties. The defaults given for each property below represent a configuration that can be implemented over TCP. An alternate set of default Protocol Selection Properties would represent a configuration that can be implemented over UDP.¶
This property specifies whether the application needs to use a transport protocol that ensures that all data is received on the other side without corruption. This also entails being notified when a Connection is closed or aborted.¶
This property specifies whether the application needs or prefers to use a transport protocol that preserves message boundaries.¶
This property specifies whether an application considers it useful to indicate its reliability requirements on a per-Message basis. This property applies to Connections and Connection Groups.¶
This property specifies whether the application wishes to use a transport protocol that can ensure that data is received by the application on the other end in the same order as it was sent.¶
This property specifies whether an application would like to supply a Message to the transport protocol before Connection establishment, which will then be reliably transferred to the other side before or during Connection establishment, potentially multiple times (i.e., multiple copies of the message data may be passed to the Remote Endpoint). See also Section 7.5.4. Note that disabling this property has no effect for protocols that are not connection-oriented and do not protect against duplicated messages, e.g., UDP.¶
This property specifies that the application would prefer multiple Connections within a Connection Group to be provided by streams of a single underlying transport connection where possible.¶
This property specifies whether the application desires protection against corruption for all data transmitted on this Connection. Disabling this property may enable to control checksum coverage later (see Section 7.5.6).¶
This property specifies whether the application desires protection against corruption for all data received on this Connection.¶
This property specifies whether the application would like the Connection to be congestion controlled or not. Note that if a Connection is not congestion controlled, an application using such a Connection should itself perform congestion control in accordance with [RFC2914]. Also note that reliability is usually combined with congestion control in protocol implementations, rendering "reliable but not congestion controlled" a request that is unlikely to succeed.¶
This property allows the application to select which specific network interfaces
or categories of interfaces it wants to Require
, Prohibit
, Prefer
, or
Avoid
. Note that marking a specific interface as Require
strictly limits path
selection to a single interface, and may often lead to less flexible and resilient
connection establishment.¶
In contrast to other Selection Properties, this property is a tuple of an (Enumerated) interface identifier and a preference, and can either be implemented directly as such, or for making one preference available for each interface and interface type available on the system.¶
The set of valid interface types is implementation- and system-specific. For
example, on a mobile device, there may be Wi-Fi
and Cellular
interface types
available; whereas on a desktop computer, there may be Wi-Fi
and Wired
Ethernet
interface types available. An implementation should provide all types
that are supported on the local system to all remote systems, to allow
applications to be written generically. For example, if a single implementation
is used on both mobile devices and desktop devices, it should define the
Cellular
interface type for both systems, since an application may want to
always Prohibit Cellular
. Note that marking a specific interface type as
Require
limits path selection to a small set of interfaces, and leads to less
flexible and resilient connection establishment.¶
The set of interface types is expected to change over time as new access technologies become available.¶
Interface types should not be treated as a proxy for properties of interfaces such as metered or unmetered network access. If an application needs to prohibit metered interfaces, this should be specified via Provisioning Domain attributes (see Section 5.2.11) or another specific property.¶
Similar to interface instances and types (see Section 5.2.10), this property
allows the application to control path selection by selecting which specific
Provisioning Domains or categories of Provisioning Domains it wants to
Require
, Prohibit
, Prefer
, or Avoid
. Provisioning Domains define
consistent sets of network properties that may be more specific than network
interfaces [RFC7556].¶
As with interface instances and types, this property is a tuple of an (Enumerated) PvD identifier and a preference, and can either be implemented directly as such, or for making one preference available for each interface and interface type available on the system.¶
The identification of a specific Provisioning Domain (PvD) is defined to be implementation- and system-specific, since there is not a portable standard format for a PvD identitfier. For example, this identifier may be a string name or an integer. As with requiring specific interfaces, requiring a specific PvD strictly limits path selection.¶
Categories or types of PvDs are also defined to be implementation- and system-specific. These may be useful to identify a service that is provided by a PvD. For example, if an application wants to use a PvD that provides a Voice-Over-IP service on a Cellular network, it can use the relevant PvD type to require some PvD that provides this service, without needing to look up a particular instance. While this does restrict path selection, it is broader than requiring specific PvD instances or interface instances, and should be preferred over these options.¶
This property allows the application to express a preference for the use of temporary local addresses, sometimes called "privacy" addresses [RFC4941]. Temporary addresses are generally used to prevent linking connections over time when a stable address, sometimes called "permanent" address, is not needed. Note that if an application Requires the use of temporary addresses, the resulting Connection cannot use IPv4, as temporary addresses do not exist in IPv4.¶
This property specifies whether an application wants to take advantage of transferring data across multiple paths between the same end hosts. Using multiple paths allows connections to migrate between interfaces as availability and performance properties change. Possible values are:¶
Enumeration values other than "Disabled" are interpreted as preferences.¶
This property specifies whether an application wants to use the connection for sending and/or receiving data. Possible values are:¶
Since unidirectional communication can be supported by transports offering bidirectional communication, specifying unidirectional communication may cause a transport stack that supports bidirectional communication to be selected.¶
This property specifies whether an application considers it useful to be informed in case sent data was retransmitted more often than a certain threshold (see Section 10.1.1 for configuration of this threshold).¶
This property specifies whether an application considers it useful to be informed when an ICMP error message arrives that does not force termination of a connection. When set to true, received ICMP errors will be available as SoftErrors, see Section 10.3.1. Note that even if a protocol supporting this property is selected, not all ICMP errors will necessarily be delivered, so applications cannot rely on receiving them.¶
Most security parameters, e.g., TLS ciphersuites, local identity and private key, etc., may be configured statically. Others are dynamically configured during connection establishment. Thus, we partition security parameters and callbacks based on their place in the lifetime of connection establishment. Similar to Transport Properties, both parameters and callbacks are inherited during cloning (see Section 6.4).¶
Common parameters such as TLS ciphersuites are known to implementations. Clients should use common safe defaults for these values whenever possible. However, as discussed in [I-D.ietf-taps-transport-security], many transport security protocols require specific security parameters and constraints from the client at the time of configuration and actively during a handshake. These configuration parameters are created as follows:¶
SecurityParameters := NewSecurityParameters()¶
Security configuration parameters and sample usage follow:¶
SecurityParameters.AddIdentity(identity) SecurityParameters.AddPrivateKey(privateKey, publicKey)¶
SecurityParameters.AddSupportedGroup(secp256k1) SecurityParameters.AddCiphersuite(TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256) SecurityParameters.AddSignatureAlgorithm(ed25519)¶
SecurityParameters.SetSessionCacheCapacity(MAX_CACHE_ELEMENTS) SecurityParameters.SetSessionCacheLifetime(SECONDS_PER_DAY) SecurityParameters.SetSessionCachePolicy(CachePolicyOneTimeUse)¶
SecurityParameters.AddPreSharedKey(key, identity)¶
Security decisions, especially pertaining to trust, are not static. Once configured, parameters may also be supplied during connection establishment. These are best handled as client-provided callbacks. Security handshake callbacks that may be invoked during connection establishment include:¶
TrustCallback := NewCallback({ // Handle trust, return the result }) SecurityParameters.SetTrustVerificationCallback(trustCallback)¶
ChallengeCallback := NewCallback({ // Handle challenge }) SecurityParameters.SetIdentityChallengeCallback(challengeCallback)¶
Before a Connection can be used for data transfer, it must be established. Establishment ends the pre-establishment phase; all transport properties and cryptographic parameter specification must be complete before establishment, as these will be used to select candidate Paths and Protocol Stacks for the Connection. Establishment may be active, using the Initiate() Action; passive, using the Listen() Action; or simultaneous for peer-to-peer, using the Rendezvous() Action. These Actions are described in the subsections below.¶
Active open is the Action of establishing a Connection to a Remote Endpoint presumed to be listening for incoming Connection requests. Active open is used by clients in client-server interactions. Active open is supported by this interface through the Initiate Action:¶
Connection := Preconnection.Initiate(timeout?)¶
The timeout parameter specifies how long to wait before aborting Active open. Before calling Initiate, the caller must have populated a Preconnection Object with a Remote Endpoint specifier, optionally a Local Endpoint specifier (if not specified, the system will attempt to determine a suitable Local Endpoint), as well as all properties necessary for candidate selection.¶
The Initiate() Action returns a Connection object. Once Initiate() has been called, any changes to the Preconnection MUST NOT have any effect on the Connection. However, the Preconnection can be reused, e.g., to Initiate another Connection.¶
Once Initiate is called, the candidate Protocol Stack(s) may cause one or more candidate transport-layer connections to be created to the specified remote endpoint. The caller may immediately begin sending Messages on the Connection (see Section 7) after calling Initate(); note that any idempotent data sent while the Connection is being established may be sent multiple times or on multiple candidates.¶
The following Events may be sent by the Connection after Initiate() is called:¶
Connection -> Ready<>¶
The Ready Event occurs after Initiate has established a transport-layer connection on at least one usable candidate Protocol Stack over at least one candidate Path. No Receive Events (see Section 8) will occur before the Ready Event for Connections established using Initiate.¶
Connection -> InitiateError<reason?>¶
An InitiateError occurs either when the set of transport properties and security parameters cannot be fulfilled on a Connection for initiation (e.g. the set of available Paths and/or Protocol Stacks meeting the constraints is empty) or reconciled with the local and/or remote Endpoints; when the remote specifier cannot be resolved; or when no transport-layer connection can be established to the remote Endpoint (e.g. because the remote Endpoint is not accepting connections, the application is prohibited from opening a Connection by the operating system, or the establishment attempt has timed out for any other reason).¶
See also Section 7.8 to combine Connection establishment and transmission of the first message in a single action.¶
Passive open is the Action of waiting for Connections from remote Endpoints, commonly used by servers in client-server interactions. Passive open is supported by this interface through the Listen Action and returns a Listener object:¶
Listener := Preconnection.Listen()¶
Before calling Listen, the caller must have initialized the Preconnection during the pre-establishment phase with a Local Endpoint specifier, as well as all properties necessary for Protocol Stack selection. A Remote Endpoint may optionally be specified, to constrain what Connections are accepted.¶
The Listen() Action returns a Listener object. Once Listen() has been called, any changes to the Preconnection MUST NOT have any effect on the Listener. The Preconnection can be disposed of or reused, e.g., to create another Listener.¶
Listening continues until the global context shuts down, or until the Stop action is performed on the Listener object:¶
Listener.Stop()¶
After Stop() is called, the Listener can be disposed of.¶
Listener -> ConnectionReceived<Connection>¶
The ConnectionReceived Event occurs when a Remote Endpoint has established a transport-layer connection to this Listener (for Connection-oriented transport protocols), or when the first Message has been received from the Remote Endpoint (for Connectionless protocols), causing a new Connection to be created. The resulting Connection is contained within the ConnectionReceived Event, and is ready to use as soon as it is passed to the application via the event.¶
Listener.SetNewConnectionLimit(value)¶
If the caller wants to rate-limit the number of inbound Connections that will be delivered, it can set a cap using SetNewConnectionLimit(). This mechanism allows a server to protect itself from being drained of resources. Each time a new Connection is delivered by the ConnectionReceived Event, the value is automatically decremented. Once the value reaches zero, no further Connections will be delivered until the caller sets the limit to a higher value. By default, this value is Infinite. The caller is also able to reset the value to Infinite at any point.¶
Listener -> ListenError<reason?>¶
A ListenError occurs either when the Properties and Security Parameters of the Preconnection cannot be fulfilled for listening or cannot be reconciled with the Local Endpoint (and/or Remote Endpoint, if specified), when the Local Endpoint (or Remote Endpoint, if specified) cannot be resolved, or when the application is prohibited from listening by policy.¶
Listener -> Stopped<>¶
A Stopped Event occurs after the Listener has stopped listening.¶
Simultaneous peer-to-peer Connection establishment is supported by the Rendezvous() Action:¶
Preconnection.Rendezvous()¶
The Preconnection Object must be specified with both a Local Endpoint and a Remote Endpoint, and also the transport properties and security parameters needed for Protocol Stack selection.¶
The Rendezvous() Action causes the Preconnection to listen on the Local Endpoint for an incoming Connection from the Remote Endpoint, while simultaneously trying to establish a Connection from the Local Endpoint to the Remote Endpoint. This corresponds to a TCP simultaneous open, for example.¶
The Rendezvous() Action returns a Connection object. Once Rendezvous() has been called, any changes to the Preconnection MUST NOT have any effect on the Connection. However, the Preconnection can be reused, e.g., for Rendezvous of another Connection.¶
Preconnection -> RendezvousDone<Connection>¶
The RendezvousDone<> Event occurs when a Connection is established with the Remote Endpoint. For Connection-oriented transports, this occurs when the transport-layer connection is established; for Connectionless transports, it occurs when the first Message is received from the Remote Endpoint. The resulting Connection is contained within the RendezvousDone<> Event, and is ready to use as soon as it is passed to the application via the Event.¶
Preconnection -> RendezvousError<reason?>¶
An RendezvousError occurs either when the Properties and Security Parameters of the Preconnection cannot be fulfilled for rendezvous or cannot be reconciled with the Local and/or Remote Endpoints, when the Local Endpoint or Remote Endpoint cannot be resolved, when no transport-layer connection can be established to the Remote Endpoint, or when the application is prohibited from rendezvous by policy.¶
When using some NAT traversal protocols, e.g., Interactive Connectivity Establishment (ICE) [RFC5245], it is expected that the Local Endpoint will be configured with some method of discovering NAT bindings, e.g., a Session Traversal Utilities for NAT (STUN) server. In this case, the Local Endpoint may resolve to a mixture of local and server reflexive addresses. The Resolve() action on the Preconnection can be used to discover these bindings:¶
[]Preconnection := Preconnection.Resolve()¶
The Resolve() call returns a list of Preconnection Objects, that represent the concrete addresses, local and server reflexive, on which a Rendezvous() for the Preconnection will listen for incoming Connections. These resolved Preconnections will share all other Properties with the Preconnection from which they are derived, though some Properties may be made more-specific by the resolution process. This list can be passed to a peer via a signalling protocol, such as SIP [RFC3261] or WebRTC [RFC7478], to configure the remote.¶
Entangled Connections can be created using the Clone Action:¶
Connection := Connection.Clone()¶
Calling Clone on a Connection yields a group of two Connections: the parent Connection on which Clone was called, and the resulting cloned Connection. These connections are "entangled" with each other, and become part of a Connection Group. Calling Clone on any of these two Connections adds a third Connection to the Connection Group, and so on. Connections in a Connection Group generally share Connection Properties. However, there may be exceptions, such as "Priority (Connection)", see Section 10.1.3. Like all other Properties, Priority is copied to the new Connection when calling Clone(), but it is not entangled: Changing Priority on one Connection does not change it on the other Connections in the same Connection Group.¶
In addition, incoming entangled Connections can be received by creating a Listener on an existing connection:¶
Listener := Connection.ListenClone()¶
ListenClone() creates a Listener that listens on the same LocalEndpoint as the one the cloned Connection is using. Any new Connection received by this Listener will be entangled with the cloned Connection. Changing one of the Connection Properties on one Connection in the group changes it for all others. Message Properties, however, are not entangled. For example, changing "Timeout for aborting Connection" (see Section 10.1.4) on one Connection in a group will automatically change this Connection Property for all Connections in the group in the same way. However, changing "Lifetime" (see Section 7.5.1) of a Message will only affect a single Message on a single Connection, entangled or not.¶
If the underlying protocol supports multi-streaming, it is natural to use this functionality to implement Clone. In that case, entangled Connections are multiplexed together, giving them similar treatment not only inside endpoints but also across the end-to-end Internet path.¶
Note that calling Clone() may result in on-the-wire signaling, e.g., to open a new connection, depending on the underlying Protocol Stack. When Clone() leads to multiple connections being opened instead of multi-streaming, the transport system will ensure consistency of Connection Properties by uniformly applying them to all underlying connections in a group. Even in such a case, there are possibilities for a transport system to implement prioritization within a Connection Group [TCP-COUPLING] [RFC8699].¶
Attempts to clone a Connection can result in a CloneError:¶
Connection -> CloneError<reason?>¶
The Connection Property "Priority" operates on entangled Connections as in Section 7.5.2: when allocating available network capacity among Connections in a Connection Group, sends on Connections with higher Priority values will be prioritized over sends on Connections with lower Priority values. A transport system implementation should, if possible, assign each Connection the capacity share (M-N) x C / M, where N is the Connection's Priority value, M is the maximum Priority value used by all Connections in the group and C is the total available capacity. However, the Priority setting is purely advisory, and no guarantees are given about the way capacity is shared. Each implementation is free to implement a way to share capacity that it sees fit.¶
Once a Connection has been established, it can be used for sending data. Data is sent as Messages, which allow the application to communicate the boundaries of the data being transferred. By default, Send enqueues a complete Message, and takes optional per-Message properties (see Section 7.1). All Send actions are asynchronous, and deliver events (see Section 7.3). Sending partial Messages for streaming large data is also supported (see Section 7.6).¶
Messages are sent on a Connection using the Send action:¶
Connection.Send(messageData, messageContext?, endOfMessage?)¶
where messageData is the data object to send.¶
The optional messageContext parameter allows adding Message Properties as described in Section 7.5. Moreover, the messageContext can be used to identify Send Events related to a specific Message (see Section 7.3) or to inspect meta-data related to the Message sent (see Section 7.4).¶
The optional endOfMessage parameter supports partial sending and is described in Section 7.6.¶
Framers can be used to extend or modify the message data with additional information that can be processed at the receiver to detect message boundaries. This is further decribed in Section 9.¶
The most basic form of sending on a connection involves enqueuing a single Data block as a complete Message, with default Message Properties. Message data is transferred as an array of bytes, and the resulting object contains both the byte array and the length of the array.¶
messageData := "hello".bytes() Connection.Send(messageData)¶
The interpretation of a Message to be sent is dependent on the implementation, and on the constraints on the Protocol Stacks implied by the Connection's transport properties. For example, a Message may be a single datagram for UDP Connections; or an HTTP Request for HTTP Connections.¶
Some transport protocols can deliver arbitrarily sized Messages, but other protocols constrain the maximum Message size. Applications can query the Connection Property "Maximum Message size on send" (Section 10.1.8.3) to determine the maximum size allowed for a single Message. If a Message is too large to fit in the Maximum Message Size for the Connection, the Send will fail with a SendError event (Section 7.3.3). For example, it is invalid to send a Message over a UDP connection that is larger than the available datagram sending size.¶
When a message is sent in response to a message received, the application may use the Message Context of the received Message to construct a Message Context for the reply.¶
replyMessageContext := requestMessageContext.reply()¶
By using the replyMessageContext
, the transport system is informed that
the message to be sent is a response and can map the response to the same underlying transport connection or stream the request was received from.
The concept of Message Contexts is described in Section 7.4.¶
Like all Actions in this interface, the Send Action is asynchronous. There are several Events that can be delivered in response to Sending a Message. Exactly one Event (Sent, Expired, or SendError) will be delivered in reponse to each call to Send.¶
Note that if partial Sends are used (Section 7.6), there will still be exactly one Send Event delivered for each call to Send. For example, if a Message expired while two requests to Send data for that Message are outstanding, there will be two Expired events delivered.¶
Connection -> Sent<messageContext>¶
The Sent Event occurs when a previous Send Action has completed, i.e., when the data derived from the Message has been passed down or through the underlying Protocol Stack and is no longer the responsibility of this interface. The exact disposition of the Message (i.e., whether it has actually been transmitted, moved into a buffer on the network interface, moved into a kernel buffer, and so on) when the Sent Event occurs is implementation-specific. The Sent Event contains a reference to the Message to which it applies.¶
Sent Events allow an application to obtain an understanding of the amount of buffering it creates. That is, if an application calls the Send Action multiple times without waiting for a Sent Event, it has created more buffer inside the transport system than an application that always waits for the Sent Event before calling the next Send Action.¶
Connection -> Expired<messageContext>¶
The Expired Event occurs when a previous Send Action expired before completion; i.e. when the Message was not sent before its Lifetime (see Section 7.5.1) expired. This is separate from SendError, as it is an expected behavior for partially reliable transports. The Expired Event contains a reference to the Message to which it applies.¶
Connection -> SendError<messageContext, reason?>¶
A SendError occurs when a Message could not be sent due to an error condition: an attempt to send a Message which is too large for the system and Protocol Stack to handle, some failure of the underlying Protocol Stack, or a set of Message Properties not consistent with the Connection's transport properties. The SendError contains a reference to the Message to which it applies.¶
Using the MessageContext object, the application can set and retrieve meta-data of the message, including Message Properties (see Section 7.5) and framing meta-data (see Section 9.2). Therefore, a MessageContext object can be passed to the Send action and is returned by each Send and Receive related event.¶
Message Properties can be set and queried using the Message Context:¶
MessageContext.add(scope?, parameter, value) PropertyValue := MessageContext.get(scope?, property)¶
To get or set Message Properties, the optional scope parameter is left empty. To get or set meta-data for a Framer, the application has to pass a reference to this Framer as the scope parameter.¶
For MessageContexts returned by send events (see Section 7.3) and receive events (see Section 8.2), the application can query information about the local and remote endpoint:¶
RemoteEndpoint := MessageContext.GetRemoteEndpoint() LocalEndpoint := MessageContext.GetLocalEndpoint()¶
Message Contexts can also be used to send messages that are flagged as a reply to other messages, see Section 7.2 for details. If the message received was sent by the remote endpoint as a reply to an earlier message and the Protocol Stack provides this information, the MessageContext of the original request can be accessed using the Message Context of the reply:¶
RequestMessageContext := MessageContext.GetOriginalRequest()¶
Applications may need to annotate the Messages they send with extra information to control how data is scheduled and processed by the transport protocols in the Connection. Therefore a message context containing these properties can be passed to the Send Action. For other uses of the message context, see Section 7.4.¶
Note that Message Properties are per-Message, not per-Send if partial Messages are sent (Section 7.6). All data blocks associated with a single Message share properties specified in the Message Contexts. For example, it would not make sense to have the beginning of a Message expire, but allow the end of a Message to still be sent.¶
A MessageContext object contains metadata for Messages to be sent or received.¶
messageData := "hello".bytes() messageContext := NewMessageContext() messageContext.add(parameter, value) Connection.Send(messageData, messageContext)¶
The simpler form of Send, which does not take any messageContext, is equivalent to passing a default MessageContext without adding any Message Properties to it.¶
If an application wants to override Message Properties for a specific message, it can acquire an empty MessageContext Object and add all desired Message Properties to that Object. It can then reuse the same messageContext Object for sending multiple Messages with the same properties.¶
Properties may be added to a MessageContext object only before the context is used for sending. Once a messageContext has been used with a Send call, modifying any of its properties is invalid.¶
Message Properties may be inconsistent with the properties of the Protocol Stacks underlying the Connection on which a given Message is sent. For example, a Connection must provide reliability to allow setting an infinite value for the lifetime property of a Message. Sending a Message with Message Properties inconsistent with the Selection Properties of the Connection yields an error.¶
The following Message Properties are supported:¶
Lifetime specifies how long a particular Message can wait to be sent to the remote endpoint before it is irrelevant and no longer needs to be (re-)transmitted. This is a hint to the transport system - it is not guaranteed that a Message will not be sent when its Lifetime has expired.¶
Setting a Message's Lifetime to infinite indicates that the application does not wish to apply a time constraint on the transmission of the Message, but it does not express a need for reliable delivery; reliability is adjustable per Message via the "Reliable Data Transfer (Message)" property (see Section 7.5.7). The type and units of Lifetime are implementation-specific.¶
This property represents a hierarchy of priorities. It can specify the priority of a Message, relative to other Messages sent over the same Connection.¶
A Message with Priority 0 will yield to a Message with Priority 1, which will yield to a Message with Priority 2, and so on. Priorities may be used as a sender-side scheduling construct only, or be used to specify priorities on the wire for Protocol Stacks supporting prioritization.¶
Note that this property is not a per-message override of the connection Priority - see Section 10.1.3. Both Priority properties may interact, but can be used independently and be realized by different mechanisms.¶
If true, it specifies that the receiver-side transport protocol stack may only deliver the Message to the receiving application after the previous ordered Message which was passed to the same Connection via the Send Action, when such a Message exists. If false, the Message may be delivered to the receiving application out of order. This property is used for protocols that support preservation of data ordering, see Section 5.2.4, but allow out-of-order delivery for certain messages, e.g., by multiplexing independent messages onto different streams.¶
If true, it specifies that a Message is safe to send to the remote endpoint more than once for a single Send Action. It is used to mark data safe for certain 0-RTT establishment techniques, where retransmission of the 0-RTT data may cause the remote application to receive the Message multiple times.¶
Note that for protocols that do not protect against duplicated messages, e.g., UDP, all messages MUST be marked as Idempotent. In order to enable protocol selection to choose such a protocol, Idempotent MUST be added to the TransportProperties passed to the Preconnection. If such a protocol was chosen, disabling Idempotent on individual messages MUST result in a SendError.¶
If true, this Message is the last one that the application will send on a Connection. This allows underlying protocols to indicate to the Remote Endpoint that the Connection has been effectively closed in the sending direction. For example, TCP-based Connections can send a FIN once a Message marked as Final has been completely sent, indicated by marking endOfMessage. Protocols that do not support signalling the end of a Connection in a given direction will ignore this property.¶
Note that a Final Message must always be sorted to the end of a list of Messages. The Final property overrides Priority and any other property that would re-order Messages. If another Message is sent after a Message marked as Final has already been sent on a Connection, the Send Action for the new Message will cause a SendError Event.¶
This property specifies the minimum length of the section of the Message, starting from byte 0, that the application requires to be delivered without corruption due to lower layer errors. It is used to specify options for simple integrity protection via checksums. A value of 0 means that no checksum is required, and -1 means that the entire Message is protected by a checksum. Only full coverage is guaranteed, any other requests are advisory, meaning that full coverage is applied anyway.¶
When true, this property specifies that a message should be sent in such a way that the transport protocol ensures all data is received on the other side without corruption. Changing the 'Reliable Data Transfer' property on Messages is only possible for Connections that were established with the Selection Property 'Configure Per-Message Reliability' enabled. When this is not the case, changing it will generate an error. Disabling this property indicates that the transport system may disable retransmissions or other reliability mechanisms for this particular Message, but such disabling is not guaranteed.¶
This enumerated property specifies the application's preferred tradeoffs for sending this Message; it is a per-Message override of the Capacity Profile connection property (see Section 10.1.6).¶
This property specifies that a message should be sent and received as a single
packet without transport-layer segmentation or network-layer fragmentation, if possible.
Attempts to send a message with this property set with a size greater to the
transport's current estimate of its maximum transmission segment size will
result in a SendError
. When used with transports supporting this functionality
and running over IP version 4, the Don't Fragment bit will be set.¶
It is not always possible for an application to send all data associated with a Message in a single Send Action. The Message data may be too large for the application to hold in memory at one time, or the length of the Message may be unknown or unbounded.¶
Partial Message sending is supported by passing an endOfMessage boolean parameter to the Send Action. This value is always true by default, and the simpler forms of Send are equivalent to passing true for endOfMessage.¶
The following example sends a Message in two separate calls to Send.¶
messageContext := NewMessageContext() messageContext.add(parameter, value) messageData := "hel".bytes() endOfMessage := false Connection.Send(messageData, messageContext, endOfMessage) messageData := "lo".bytes() endOfMessage := true Connection.Send(messageData, messageContext, endOfMessage)¶
All data sent with the same MessageContext object will be treated as belonging to the same Message, and will constitute an in-order series until the endOfMessage is marked. Once the end of the Message is marked, the MessageContext object may be re-used as a new Message with identical parameters.¶
To reduce the overhead of sending multiple small Messages on a Connection, the application may want to batch several Send actions together. This provides a hint to the system that the sending of these Messages should be coalesced when possible, and that sending any of the batched Messages may be delayed until the last Message in the batch is enqueued.¶
Connection.Batch( Connection.Send(messageData) Connection.Send(messageData) )¶
For application-layer protocols where the Connection initiator also sends the first message, the InitiateWithSend() action combines Connection initiation with a first Message sent:¶
Connection := Preconnection.InitiateWithSend(messageData, messageContext?, timeout?)¶
Whenever possible, a messageContext should be provided to declare the message passed to InitiateWithSend as idempotent. This allows the transport system to make use of 0-RTT establishment in case this is supported by the available protocol stacks. When the selected stack(s) do not support transmitting data upon connection establishment, InitiateWithSend is identical to Initiate() followed by Send().¶
Neither partial sends nor send batching are supported by InitiateWithSend().¶
The Events that may be sent after InitiateWithSend() are equivalent to those that would be sent by an invocation of Initate() followed immediately by an invocation of Send(), with the caveat that a send failure that occurs because the Connection could not be established will not result in a SendError separate from the InitiateError signaling the failure of Connection establishment.¶
Once a Connection is established, it can be used for receiving data. As with sending, data is received in terms of Messages. Receiving is an asynchronous operation, in which each call to Receive enqueues a request to receive new data from the connection. Once data has been received, or an error is encountered, an event will be delivered to complete any pending Receive requests (see Section 8.2). If Messages arrive at the transport system before Receive requests are issued, ensuing Receive requests will first operate on these Messages before awaiting any further Messages.¶
Receive takes two parameters to specify the length of data that an application is willing to receive, both of which are optional and have default values if not specified.¶
Connection.Receive(minIncompleteLength?, maxLength?)¶
By default, Receive will try to deliver complete Messages in a single event (Section 8.2.1).¶
The application can set a minIncompleteLength value to indicate the smallest partial Message data size in bytes that should be delivered in response to this Receive. By default, this value is infinite, which means that only complete Messages should be delivered (see Section 8.2.2 and Section 9 for more information on how this is accomplished). If this value is set to some smaller value, the associated receive event will be triggered only when at least that many bytes are available, or the Message is complete with fewer bytes, or the system needs to free up memory. Applications should always check the length of the data delivered to the receive event and not assume it will be as long as minIncompleteLength in the case of shorter complete Messages or memory issues.¶
The maxLength argument indicates the maximum size of a Message in bytes the application is currently prepared to receive. The default value for maxLength is infinite. If an incoming Message is larger than the minimum of this size and the maximum Message size on receive for the Connection's Protocol Stack, it will be delivered via ReceivedPartial events (Section 8.2.2).¶
Note that maxLength does not guarantee that the application will receive that many bytes if they are available; the interface may return ReceivedPartial events with less data than maxLength according to implementation constraints. Note also that maxLength and minIncompleteLength are intended only to manage buffering, and are not interpreted as a receiver preference for message reordering.¶
Each call to Receive will be paired with a single Receive Event, which can be a success or an error. This allows an application to provide backpressure to the transport stack when it is temporarily not ready to receive messages.¶
Connection -> Received<messageData, messageContext>¶
A Received event indicates the delivery of a complete Message. It contains two objects, the received bytes as messageData, and the metadata and properties of the received Message as messageContext.¶
The messageData object provides access to the bytes that were received for this Message, along with the length of the byte array. The messageContext is provided to enable retrieving metadata about the message and referring to the message, e.g., to send replies and map responses to their requests. See Section 7.4 for details.¶
See Section 9 for handling Message framing in situations where the Protocol Stack only provides a byte-stream transport.¶
Connection -> ReceivedPartial<messageData, messageContext, endOfMessage>¶
If a complete Message cannot be delivered in one event, one part of the Message may be delivered with a ReceivedPartial event. In order to continue to receive more of the same Message, the application must invoke Receive again.¶
Multiple invocations of ReceivedPartial deliver data for the same Message by passing the same MessageContext, until the endOfMessage flag is delivered or a ReceiveError occurs. All partial blocks of a single Message are delivered in order without gaps. This event does not support delivering discontiguous partial Messages.¶
If the minIncompleteLength in the Receive request was set to be infinite (indicating a request to receive only complete Messages), the ReceivedPartial event may still be delivered if one of the following conditions is true:¶
Note that in the absence of message boundary preservation or a Message Framer, all bytes received on the Connection will be represented as one large Message of indeterminate length.¶
Connection -> ReceiveError<messageContext, reason?>¶
A ReceiveError occurs when data is received by the underlying Protocol Stack that cannot be fully retrieved or parsed, or when some other indication is received that reception has failed. In contrast, conditions that irrevocably lead to the termination of the Connection are signaled using ConnectionError instead (see Section 11).¶
The ReceiveError event passes an optional associated MessageContext. This may indicate that a Message that was being partially received previously, but had not completed, encountered an error and will not be completed.¶
Each Message Context may contain metadata from protocols in the Protocol Stack; which metadata is available is Protocol Stack dependent. These are exposed though additional read-only Message Properties that can be queried from the MessageContext object (see Section 7.4) passed by the receive event. The following metadata values are supported:¶
When available, Message metadata carries the value of the Explicit Congestion Notification (ECN) field. This information can be used for logging and debugging purposes, and for building applications which need access to information about the transport internals for their own operation. This property is specific to UDP and UDP-Lite because these protocols do not implement congestion control, and hence expose this functionality to the application.¶
In some cases it may be valuable to know whether data was read as part of early data transfer (before connection establishment has finished). This is useful if applications need to treat early data separately, e.g., if early data has different security properties than data sent after connection establishment. In the case of TLS 1.3, client early data can be replayed maliciously (see [RFC8446]). Thus, receivers may wish to perform additional checks for early data to ensure it is idempotent or not replayed. If TLS 1.3 is available and the recipient Message was sent as part of early data, the corresponding metadata carries a flag indicating as such. If early data is enabled, applications should check this metadata field for Messages received during connection establishment and respond accordingly.¶
The Message Context can indicate whether or not this Message is the Final Message on a Connection. For any Message that is marked as Final, the application can assume that there will be no more Messages received on the Connection once the Message has been completely delivered. This corresponds to the Final property that may be marked on a sent Message, see Section 7.5.5.¶
Some transport protocols and peers may not support signaling of the Final property. Applications therefore should not rely on receiving a Message marked Final to know that the other endpoint is done sending on a connection.¶
Any calls to Receive once the Final Message has been delivered will result in errors.¶
Although most applications communicate over a network using well-formed Messages, the boundaries and metadata of the Messages are often not directly communicated by the transport protocol itself. For example, HTTP applications send and receive HTTP messages over a byte-stream transport, requiring that the boundaries of HTTP messages be parsed out from the stream of bytes.¶
Message Framers allow extending a Connection's Protocol Stack to define how to encapsulate or encode outbound Messages, and how to decapsulate or decode inbound data into Messages. Message Framers allow message boundaries to be preserved when using a Connection object, even when using byte-stream transports. This facility is designed based on the fact that many of the current application protocols evolved over TCP, which does not provide message boundary preservation, and since many of these protocols require message boundaries to function, each application layer protocol has defined its own framing.¶
Note that while Message Framers add the most value when placed above a protocol that otherwise does not preserve message boundaries, they can also be used with datagram- or message-based protocols. In these cases, they add an additional transformation to further encode or encapsulate, and can potentially support packing multiple application-layer Messages into individual transport datagrams.¶
The API to implement a Message Framer can vary depending on the implementation; guidance on implementing Message Framers can be found in [I-D.ietf-taps-impl].¶
The Message Framer object can be added to one or more Preconnections to run on top of transport protocols. Multiple Framers may be added. If multiple Framers are added, the last one added runs first when framing outbound messages, and last when parsing inbound data.¶
The following example adds a basic HTTP Message Framer to a Preconnection:¶
framer := NewHTTPMessageFramer() Preconnection.AddFramer(framer)¶
When sending Messages, applications can add specific Message values to a MessageContext (Section 7.4) that is intended for a Framer. This can be used, for example, to set the type of a Message for a TLV format. The namespace of values is custom for each unique Message Framer.¶
messageContext := NewMessageContext() messageContext.add(framer, key, value) Connection.Send(messageData, messageContext)¶
When an application receives a MessageContext in a Receive event, it can also look to see if a value was set by a specific Message Framer.¶
messageContext.get(framer, key) -> value¶
For example, if an HTTP Message Framer is used, the values could correspond to HTTP headers:¶
httpFramer := NewHTTPMessageFramer() ... messageContext := NewMessageContext() messageContext.add(httpFramer, "accept", "text/html")¶
During pre-establishment and after establishment, connections can be configured and queried using Connection Properties, and asynchronous information may be available about the state of the connection via Soft Errors.¶
Connection Properties represent the configuration and state of the selected Protocol Stack(s) backing a Connection. These Connection Properties may be Generic, applying regardless of transport protocol, or Specific, applicable to a single implementation of a single transport protocol stack. Generic Connection Properties are defined in Section 10.1 below. Specific Protocol Properties are defined in a transport- and implementation-specific way, and must not be assumed to apply across different protocols. Attempts to set Specific Protocol Properties on a protocol stack not containing that specific protocol are simply ignored, and do not raise an error; however, too much reliance by an application on Specific Protocol Properties may significantly reduce the flexibility of a transport services implementation.¶
The application can set and query Connection Properties on a per-Connection basis. Connection Properties that are not read-only can be set during pre-establishment (see Section 5.2), as well as on connections directly using the SetProperty action:¶
Connection.SetProperty(property, value)¶
Note that changing one of the Connection Properties on one Connection in a Connection Group will also change it for all other Connections of that group; see further Section 6.4.¶
At any point, the application can query Connection Properties.¶
ConnectionProperties := Connection.GetProperties()¶
Depending on the status of the connection, the queried Connection Properties will include different information:¶
Generic Connection Properties are defined independent of the chosen protocol stack and therefore available on all Connections.¶
Note that many Connection Properties have a corresponding Selection Property which enables applications to express their preference for protocols providing a supporting transport feature.¶
This property specifies after how many retransmissions to inform the application about "Excessive Retransmissions". The special value -1 means that this notification is disabled.¶
This property specifies the part of the received data that needs to be covered by a checksum. It is given in Bytes. A value of 0 means that no checksum is required, and the special value -1 indicates full checksum coverage.¶
This Property is a non-negative integer representing the relative inverse priority (i.e., a lower value reflects a higher priority) of this Connection relative to other Connections in the same Connection Group. It has no effect on Connections not part of a Connection Group. As noted in Section 6.4, this property is not entangled when Connections are cloned, i.e., changing the Priority on one Connection in a Connection Group does not change it on the other Connections in the same Connection Group.¶
This property specifies how long to wait before deciding that a Connection has failed when trying to reliably deliver data to the destination. Adjusting this Property will only take effect when the underlying stack supports reliability. The special value -1 means that this timeout is not scheduled to happen. This can be a valid choice with unreliable data transfer (e.g., when UDP is the underlying transport protocol).¶
This property specifies which scheduler should be used among Connections within a Connection Group, see Section 6.4. The set of schedulers can be taken from [RFC8260].¶
This property specifies the desired network treatment for traffic sent by the application and the tradeoffs the application is prepared to make in path and protocol selection to receive that desired treatment. When the capacity profile is set to a value other than Default, the transport system SHOULD select paths and configure protocols to optimize the tradeoff between delay, delay variation, and bandwidth efficiency based on the capacity profile specified. How this is realized is implementation-specific. The Capacity Profile MAY also be used to set priorities on the wire for Protocol Stacks supporting prioritization. Recommendations for use with DSCP are provided below for each profile; note that when a Connection is multiplexed, the guidelines in Section 6 of [RFC7657] apply.¶
The following values are valid for the Capacity Profile:¶
The Capacity Profile for a selected protocol stack may be modified on a per-Message basis using the Transmission Profile Message Property; see Section 7.5.8.¶
This property specifies an upper-bound rate that a transfer is not expected to exceed (even if flow control and congestion control allow higher rates), and/or a lower-bound rate below which the application does not deem a data transfer useful. It is given in bits per second. The special value -1 indicates that no bound is specified.¶
The following generic Connection Properties are read-only, i.e. they cannot be changed by an application.¶
This property represents the maximum Message size that can be sent before or during Connection establishment, see also Section 7.5.4. It is given in Bytes.¶
This property, if applicable, represents the maximum Message size that can be sent without incurring network-layer fragmentation or transport layer segmentation at the sender. This property exposes the Maximum Packet Size (MPS) as described in Datagram PLPMTUD [I-D.ietf-tsvwg-datagram-plpmtud].¶
This property represents the maximum Message size that can be sent using a send operation.¶
This numeric property represents the maximum Message size that can be received.¶
These properties specify configurations for the User Timeout Option (UTO), in case TCP becomes the chosen transport protocol. Implementation is optional and of course only sensible if TCP is implemented in the transport system.¶
These TCP-specific properties are included here because the feature "Suggest timeout to the peer" is part of the minimal set of transport services [I-D.ietf-taps-minset], where this feature was categorized as "functional". This means that when an implementation offers this feature, it has to expose an interface to it to the application. Otherwise, the implementation might violate assumptions by the application, which could cause the application to fail.¶
All of the below properties are optional (e.g., it is possible to specify "User Timeout Enabled" as true, but not specify an Advertised User Timeout value; in this case, the TCP default will be used).¶
This time value is advertised via the TCP User Timeout Option (UTO) [RFC5482] at the remote endpoint to adapt its own "Timeout for aborting Connection" (see Section 10.1.4) value accordingly.¶
This property controls whether the UTO option is enabled for a connection. This applies to both sending and receiving.¶
This property controls whether the "Timeout for aborting Connection" (see Section 10.1.4) may be changed based on a UTO option received from the remote peer. This boolean becomes false when "Timeout for aborting Connection" (see Section 10.1.4) is used.¶
During the lifetime of a connection there are events that can occur when configured.¶
Asynchronous introspection is also possible, via the SoftError Event. This event informs the application about the receipt and contents of an ICMP error message related to the Connection. This will only happen if the underlying protocol stack supports access to soft errors; however, even if the underlying stack supports it, there is no guarantee that a soft error will be signaled.¶
Connection -> SoftError<>¶
This event notifies the application of excessive retransmissions, based on a configured threshold (see Section 10.1.1). This will only happen if the underlying protocol stack supports reliability and, with it, such notifications.¶
Connection -> ExcessiveRetransmission<>¶
Close terminates a Connection after satisfying all the requirements that were specified regarding the delivery of Messages that the application has already given to the transport system. For example, if reliable delivery was requested for a Message handed over before calling Close, the transport system will ensure that this Message is indeed delivered. If the Remote Endpoint still has data to send, it cannot be received after this call.¶
Connection.Close()¶
The Closed Event can inform the application that the Remote Endpoint has closed the Connection; however, there is no guarantee that a remote Close will indeed be signaled.¶
Connection -> Closed<>¶
Abort terminates a Connection without delivering remaining data:¶
Connection.Abort()¶
A ConnectionError informs the application that data to could not be delivered after a timeout, or the other side has aborted the Connection; however, there is no guarantee that an Abort will indeed be signaled.¶
Connection -> ConnectionError<reason?>¶
As this interface is designed to be independent of an implementation's concurrency model, the details of how exactly actions are handled, and how events are dispatched, are implementation dependent.¶
Each transition of connection state is associated with one of more events:¶
The interface provides the following guarantees about the ordering of operations:¶
RFC-EDITOR: Please remove this section before publication.¶
This document has no Actions for IANA. Later versions of this document may create IANA registries for generic transport property names and transport property namespaces (see Section 4.2.1).¶
This document describes a generic API for interacting with a transport services (TAPS) system. Part of this API includes configuration details for transport security protocols, as discussed in Section 5.3. It does not recommend use (or disuse) of specific algorithms or protocols. Any API-compatible transport security protocol should work in a TAPS system. Security consideration for these protocols should be discussed in the respective specifications.¶
The desribed API is used to exchange information between an application and the transport system. While it is not necessarily expected that both systems are implemented by the same authority, it is expected that the transport system implementation is either provided as a library that is selected by the application from a trusted party, or that it is part of the operating system that the application also relies on for other tasks.¶
In either case, the TAPS API is an internal interface that is used to change information locally between two systems. However, as the transport system is responsible for network communication, it is in the position to potentially share any information provided by the application with the network or another communication peer. Most of the information provided over the TAPS API are useful to configure and select protocols and paths and are not necessarily privacy sensitive. Still, there is some information that could be privacy sensitve because this might reveal usage characteristics and habits of the user of an application.¶
Of course any communication over a network reveals usage characteristics, as all packets as well as their timing and size are part of the network-visible wire image [RFC8546]. However, the selection of a protocol and its configuration also impacts which information is visible, potentially in clear text, and which other enties can access it. In most cases information that is provided for protocol and path selection should not directly translate to information that is can be observed by network devices on the path. But there might be specific configuration information that are intended for path exposure, such as e.g. a DiffServ codepoint setting, that is either povided directly by the application or indirectly configured over a traffic profile.¶
Further, applications should be aware that communication attempts can lead to more than one connection establishment. This is for example the case when the transport system also excecutes name resolution; or when support mechanisms such as TURN or ICE are used to establish connectivity; or if protocols or paths are raised; or if a path fails and fallback or re-establishment is supported in the transport system.¶
These communication activities are not different from what is used today, however, the goal of a TAPS transport system is to support such mechanisms as a generic service within the transport layer. This enables applications to more dynamically benefit from innovations and new protocols in the transport system but at the same time may reduce transparency of the underlying communication actions to the application itself. The TAPS API is designed such that protocol and path selection can be limited to a small and controlled set if required by the application for functional or security purposes. Further, TAPS implementations should provide an interface to poll information about which protocol and path is currently in use as well as provide logging about the communication events of each connection.¶
This work has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreements No. 644334 (NEAT) and No. 688421 (MAMI).¶
This work has been supported by Leibniz Prize project funds of DFG - German Research Foundation: Gottfried Wilhelm Leibniz-Preis 2011 (FKZ FE 570/4-1).¶
This work has been supported by the UK Engineering and Physical Sciences Research Council under grant EP/R04144X/1.¶
This work has been supported by the Research Council of Norway under its "Toppforsk" programme through the "OCARINA" project.¶
Thanks to Stuart Cheshire, Josh Graessley, David Schinazi, and Eric Kinnear for their implementation and design efforts, including Happy Eyeballs, that heavily influenced this work. Thanks to Laurent Chuat and Jason Lee for initial work on the Post Sockets interface, from which this work has evolved. Thanks to Maximilian Franke for asking good questions based on implementation experience and for contributing text, e.g., on multicast.¶
As Selection Properties of type Preference
will be added to a TransportProperties object quite frequently, implementations should provide special actions for adding each preference level i.e, TransportProperties.Add(some_property, avoid)
is equivalent to TransportProperties.Avoid(some_property)
:¶
TransportProperties.Require(property) TransportProperties.Prefer(property) TransportProperties.Ignore(property) TransportProperties.Avoid(property) TransportProperties.Prohibit(property) TransportProperties.Default(property)¶
To ease the use of the interface specified by this document, implementations should provide a mechanism to create Transport Property objects (see Section 5.2) that are pre-configured with frequently used sets of properties. Implementations should at least offer short-hands to specify the following property profiles:¶
This profile provides reliable, in-order transport service with congestion control. An example of a protocol that provides this service is TCP. It should consist of the following properties:¶
Property | Value |
---|---|
reliability | require |
preserve-order | require |
congestion-control | require |
preserve-msg-boundaries | ignore |
This profile provides message-preserving, reliable, in-order transport service with congestion control. An example of a protocol that provides this service is SCTP. It should consist of the following properties:¶
Property | Value |
---|---|
reliability | require |
preserve-order | require |
congestion-control | require |
preserve-msg-boundaries | require |
This profile provides unreliable datagram transport service. An example of a protocol that provides this service is UDP. It should consist of the following properties:¶
Property | Value |
---|---|
reliability | ignore |
preserve-order | ignore |
congestion-control | ignore |
preserve-msg-boundaries | require |
idempotent | true |
Applications that choose this Transport Property Profile for latency reasons should also consider setting the Capacity Profile Property, see Section 10.1.6 accordingly and my benefit from controlling checksum coverage, see Section 5.2.7 and Section 5.2.8.¶
[I-D.ietf-taps-minset] identifies a minimal set of transport services that end systems should offer. These services make all non-security-related transport features of TCP, MPTCP, UDP, UDP-Lite, SCTP and LEDBAT available that 1) require interaction with the application, and 2) do not get in the way of a possible implementation over TCP (or, with limitations, UDP). The following text explains how this minimal set is reflected in the present API. For brevity, it is based on the list in Section 4.1 of [I-D.ietf-taps-minset], updated according to the discussion in Section 5 of [I-D.ietf-taps-minset]. This list is a subset of the transport features in Appendix A of [I-D.ietf-taps-minset], which refers to the primitives in "pass 2" (Section 4) of [RFC8303] for further details on the implementation with TCP, MPTCP, UDP, UDP-Lite, SCTP and LEDBAT.¶
Initiate
(Section 6.1) or InitiateWithSend
Action (Section 7.8).¶
Parallel Use of Multiple Paths
Property (Section 5.2.13).¶
InitiateWithSend
Action (Section 7.8).¶
Timeout for Aborting Connection
property, using a time value (Section 10.1.4).¶
ConnectionError
Event (Section 11).¶
TCP-specific Property: User Timeout
(Section 10.2).¶
Notification of excessive retransmissions
property (Section 5.2.15).¶
Notification of ICMP soft error message arrival
property (Section 5.2.16).¶
Connection Group Transmission Scheduler
property (Section 10.1.5).¶
Priority (Connection)
property (Section 10.1.3).¶
Corruption Protection Length
property (Section 7.5.6) and Full Checksum Coverage on Sending
property (Section 5.2.7).¶
Required Minimum Corruption Protection Coverage for Receiving
property (Section 10.1.2) and Full Checksum Coverage on Receiving
property (Section 5.2.8).¶
Singular Transmission
Message Property combines both of these requests, i.e. if a request not to bundle messages is made, this also turns off fragmentation (i.e., sets DF=1) in case of protocols that allow this (only UDP and UDP-Lite, which cannot bundle messages anyway) (Section 7.5.9).¶
Maximum Message Size Before Fragmentation or Segmentation
property (Section 10.1.8.2).¶
Maximum Message Size on Receive
property (Section 10.1.8.4).¶
ECN
is a defined UDP(-Lite)-specific read-only Message Property of the MessageContext object (Section 8.3.1).¶
Capacity Profile
property (Section 10.1.6). Per-Message control is offered via the Message Capacity Profile Override
property (Section 7.5.8).¶
Close
Action with slightly changed semantics in line with the discussion in Section 5.2 of [I-D.ietf-taps-minset] (Section 11).¶
Abort
action without promising that this is signaled to the other side. If it is, a ConnectionError
Event will fire at the peer (Section 11).¶
Send
action (Section 7). Reliability is controlled via the Reliable Data Transfer (Connection)
(Section 5.2.1) property and the Reliable Data Transfer (Message)
Message Property (Section 7.5.7). Transmitting data as a message or without delimiters is controlled via Message Framers (Section 9). The choice of congestion control is provided via the Congestion control
property (Section 5.2.9).¶
Lifetime
Message Property implements a time-based way to configure message reliability (Section 7.5.1).¶
Ordered
(Section 7.5.3).¶
Capacity Profile
Property (Section 10.1.6) or the Message Capacity Profile Override
Message Property (Section 7.5.8) with value "Low Latency/Interactive".¶
Received
Event (Section 8.2.1). See Section 9 for handling Message framing in situations where the Protocol
Stack only provides a byte-stream transport.¶
Received
Event (Section 8.2.1), using Message Framers (Section 9).¶
ReceivedPartial
Event (Section 8.2.2).¶
Expired
Event (Section 7.3.2) and SendError
Event (Section 7.3.3).¶
Sent
Event (Section 7.3.1).¶
ReceiveError
Event (Section 8.2.3).¶