In computer science, choreographic programming is a programming paradigm where programs are compositions of interactions among multiple concurrent participants. [1] [2] [3]
In choreographic programming, developers use a choreographic programming language to define the intended communication behaviour of concurrent participants. Programs in this paradigm are called choreographies. [1] Choreographic languages are inspired by security protocol notation (also known as "Alice and Bob" notation). The key to these languages is the communication primitive, for example
Alice.expr -> Bob.x reads "Alice communicates the result of evaluating the expression expr to Bob, which stores it in its local variable x". [3] Alice, Bob, etc. are typically called roles or processes. [2]
The example below shows a choreography for a simplified single sign-on (SSO) protocol based on a Central Authentication Service (CAS) that involves three roles:
Client, which wishes to obtain an access token from CAS to interact with Service.Service, which needs to know from CAS if the Client should be given access.CAS, which is the Central Authentication Service responsible for checking the Client's credentials.The choreography is:
Client.(credentials, serviceID) -> CAS.authRequest if CAS.check(authRequest) then CAS.token = genToken(authRequest) CAS.Success(token) -> Client.result CAS.Success(token) -> Service.result else CAS.Failure -> Client.result CAS.Failure -> Service.result The choreography starts in Line 1, where Client communicates a pair consisting of some credentials and the identifier of the service it wishes to access to CAS. CAS stores this pair in its local variable authRequest (for authentication request). In Line 2, the CAS checks if the request is valid for obtaining an authentication token. If so, it generates a token and communicates a Success message containing the token to both Client and Service (Lines 3–5). Otherwise, the CAS informs Client and Service that authentication failed, by sending a Failure message (Lines 7–8). We refer to this choreography as the "SSO choreography" in the remainder.
A key feature of choreographic programming is the capability of compiling choreographies to distributed implementations. These implementations can be libraries for software that needs to participate in a computer network by following a protocol, [1] [3] [4] or standalone distributed programs. [5] [6]
The translation of a choreography into distributed programs is called endpoint projection (EPP for short). [2] [3]
Endpoint projection returns a program for each role described in the source choreography. [3] For example, given the choreography above, endpoint projection would return three programs: one for Client, one for Service, and one for CAS. They are shown below in pseudocode form, where send and recv are primitives for sending and receiving messages to/from other roles.
| Client | send (credentials, serviceID) to CAS recv result from CAS |
|---|---|
| Service | recv result from CAS |
| CAS | recv authRequest from Client if check(authRequest) then token = genToken(authRequest) send Success(token) to Client send Success(token) to Service else send Failure to Client send Failure to Service |
For each role, its code contains the actions that the role should execute to implement the choreography correctly together with the others.
The paradigm of choreographic programming originates from its titular PhD thesis. [7] [8] [9] The inspiration for the syntax of choreographic programming languages can be traced back to security protocol notation, also known as "Alice and Bob" notation. [1] Choreographic programming has also been heavily influenced by standards for service choreography and interaction diagrams, as well as developments of the theory of process calculi. [1] [3] [10]
Choreographic programming is an active area of research. The paradigm has been used in the study of information flow, [11] parallel computing, [12] cyber-physical systems, [13] [14] runtime adaptation, [6] and system integration. [15]
{{cite book}}: |work= ignored (help)