Interface control document
An interface control drawing or interface control document (ICD) in systems engineering  and software engineering, describes the interface or interfaces between subsystems or to a system or subsystem.
An ICD may describe:
- the inputs and outputs of a single system, e.g. "The Wikipedia Interface Control Document."
- the interface between two systems or subsystems, e.g. "The Doghouse to Outhouse Interface Control Document."
- the complete interface protocol from the lowest physical elements (e.g., the mating plugs, the electrical signal voltage levels) to the highest logical levels (e.g., the level 7 application layer of the ISO model), or some subset thereof.
The purpose of the ICD is to communicate all possible inputs to and all potential outputs from a system for some potential or actual user of the system. The internal interfaces of a system or subsystem are typically not documented in an ICD, but rather in a system design document (such as a software design document).
Interface control documents are a key element of systems engineering as they define and control the interface(s) of a system, and thereby bound its requirements.
An application programming interface is a form of ICD for a software system, in that it describes how to access the functions and services provided by a system via an interface. If a system producer wants others to be able to use the system, an ICD (or equivalent) is a worthwhile investment.
An ICD should only describe the interface itself, and not the characteristics of the systems which use it to connect. The function and logic of those systems should be described in their own design documents if required. In this way, independent teams can develop the connecting systems which use the interface specified, without regard to how other systems will react to data and signals which are sent over the interface. For example, the ICD must include information about the size, format, and what is measured by the data, but not any ultimate meaning of the data in its intended use by any user.
An adequately defined ICD will allow one team to test its implementation of the interface by simulating the opposing side with a simple communications simulator. Not knowing the business logic of the system on the far side of an interface makes it more likely that one will develop a system that does not break when the other system changes its business rules and logic. (Provision for limits or sanity checking should be pointedly avoided in an ICD.) Thus, good modularity and abstraction leading to easy maintenance and extensibility are achieved.
Critics of ICDs and systems engineering in general often complain of the over-emphasis on documentation.  ICDs are often present on document-driven projects, but may be useful on agile projects as well (although not explicitly named as such). An ICD need not be a textual document. It may be an (evolving) table of goes-intos and comes-out-ofs, a dynamic database representing each subsystem as a DB view, a set of interaction diagrams, etc.
ICDs are often used where subsystems are developed asynchronously in time, since they provide a structured way to communicate information about subsystems interfaces between different subsystem design teams.   
The following are examples of completed ICDs or templates for ICDs.
- U.S. Government Veteran's Administration Software ICD template in Microsoft Word format
- U.S. Government Health and Human Services ICD template in PDF format
- SpaceBel Service Support Environment completed dealing with interface between Earth Observation and GIS Services; SpaceBel is a EU software engineering firm
- U.S. Government's GPS system - various ICDs completed a series of current documents, and defines Interface Specifications as a specific type of ICD where only one agency has approval responsibility
- Wolter J. Fabrycky, Benjamin S. Blanchard (2005). Systems Engineering and Analysis. Prentice-Hall, 2005
- Fowler, M.; J. Highsmith (July 2001). "The Agile Manifesto". Dr. Dobb's Journal. Retrieved 2006-05-11., "Yes, physical documentation has heft and substance, but the real measure of success is abstract: Will the people involved gain the understanding they need?"
- Ambler, S.W. (March 2005). "Agile Modeling and eXtreme Programming (XP)". AgileModeling.com. Retrieved 2006-05-11., "...verbal communication between team members reduces the need for documentation within the team."
- Agile/Lean Documentation: Strategies for Agile Software Development
- Much Ado About Nothing: Documentation
- Cutkosky, Mark R.; Jay M. Tenenbaum; Jay Glicksman (September 1996). "Madefast: collaborative engineering over the Internet". Communications of the ACM 39 (9): 78–87. doi:10.1145/234215.234474. Retrieved 2012-12-12.
- Spinellis, Diomidis (November 1998). "A Critique of the Windows Application Programming Interface". Computer Standards & Interfaces 20 (1): 1–8. Retrieved 2012-12-12.
- Leonard, Jason (May 2002). "Bringing System Engineers and Software Engineers Together". The Rational Edge. Retrieved 2012-12-12.