Problem: Assigning tasks and roles across a geographically distributed work force.
Context:
Forces:
Communication patterns between project members follows geographic distribution.
Coupling between pieces of software must be sustained by analogous coupling between the people maintaining that software.
People avoid communicating with people who work in other buildings, other towns, or overseas (see the Rationale).
People in an organization usually work on related tasks, which suggests that they communicate frequently with each other.
Solution:
The architectural partitioning should reflect the geographic partitioning, and vice versa. Architectural responsibilities should be assigned so decisions can be made (geographically) locally.
You still need someone to break logjams when consensus can't be reached, perhaps using Architect Also Implements or Patron.
If the organization is modularized along geographic boundaries, and the architecture is not, then it will be impossible to apply Architect Also Implements.
Design Rationale:
Thomas Allen [Allen] has found that social distance goes up rapidly with physical separation (see also ``House Cluster'' [Alexander 37] of Alexander). Our empirical experience with co-development projects overseas reveals that failure to follow this pattern can lead to complete project failure. This is a crucial pattern that is often overlooked or dismissed out of consideration for political alliances. Peter Bürgi's studies of geographically distributed organizations in AT&T underscore the importance of this pattern.
We have seen few geographically distributed organizations that exhibit positive team dynamics. There are exceptions, and there are rare occasions when this pattern does not apply. Steve Berczuk at MIT notes: "... communications need not be poor between remote sites if the following items are true: 1) the number of developers on a project, including all sites is small; 2) most of the communication is done via something like email (wide distribution and asynchronous communication--in [one case of his experience] ... more people were in the loop than if the primary means of communication had been hallway chats); 3) The people involved have been together for SOME time so that they feel like they know each other; 4) Folks aren't so burned out by `unnecessary' travel that they are willing and happy to travel when it is needed. In some situations pattern 7 is not possible because of the nature of the project, so we need a way to address the issue of remoteness." (Personal communication with Steve Berczuk, August, 1994.)
There are times when the market demands geographic distribution; see Organization Follows Market.
Last updated
Thu Mar 23 09:00:44 CST 1995
Copyright © 1995 AT&T