Permalink
Fetching contributors…
Cannot retrieve contributors at this time
290 lines (223 sloc) 8.76 KB

InnerSource Pattern System

About

This document shall serve to define a system with which we organize our patterns using a limited set of classifications.

Related Work

Design Patterns (Erich Gamma et. al.)

From the Authors of the classic GoF book "Design Patterns":

Design patterns vary in their granularity and level of abstraction. Because there are many design patterns, we need a way to organize them.

They classify patterns so that they "can refer to families of related patterns." The authors also suspect that this classification will make "learning the catalog" faster. They Classify patterns according to two criteria: Purpose and Scope. They introduce the following classes of purpose in their book:

  • Creational,
  • Structural and
  • Behavioral.

Orthogonal to this, they distinguish two scope classes, related to object oriented SW development:

  • Class and
  • Object.

The main ordering criterion in the books catalog is the Purpose and each pattern is "tagged" with the scope classification.

Pattern-Oriented Software Architecture (Fram Buschmann et. al.)

Frank Buschmann also organizes the patterns in his (first) book using, what he calls "categories". Similarly to the GoF book he uses one category to organize the content of his book, in this case the level of abstraction the pattern applies on:

  • Architectural Patterns,
  • Design Patterns and
  • Idioms.

The second criterion categorizes the problems solved by the patterns:

  • From Mud to Structure,
  • Distributed Systems,
  • Interactive Systems,
  • Adaptable Systems,
  • Structural Decomposition,
  • ...

you get the point.

In addition to this, he postulates a number of properties, a pattern system must exhibit. According to Buschmann, a pattern system

  • should be simple and easy to learn,
  • should consist of only a few classification criteria,
  • should use a classification criterion that the reflects natural properties of patterns (e. g. kinds of problems, not e. g. what pattern language a patterns belongs to),
  • should provide a roadmap that leads users to a set of potentially applicable patterns and
  • it should be open to the integration of new patterns.

Conclusion

I have used both books extensively and found their respective pattern systems very helpful in finding patterns for specific problems and for finding related patterns. I also think that the criterions for a useful pattern system postulated by Frank Buschmanns are sensible and might serve as a role model for our InnerSource pattern system.

A pattern language for creating pattern languages (Takashi Iba)

Takashi Iba has published an article in the ACM Digital Library from PLoP 2016: A pattern language for creating pattern languages: 364 patterns for pattern mining, writing, and symbolizing

Candiate Classifications

This section shall serve to collect individual proposals for systems of ISC patterns. Contribute away ;)

Georg Grütter

I propose we use a classification of the problem, a pattern tries to solve as the main criteria for organizing our list of patterns. The following classifications come to mind:

    1. Getting started with InnerSource
    1. Getting sustainable middle management buy-in and support
    1. Getting buy-in of developers
    1. Fostering contributions
    1. Interfacing with traditional, non-InnerSource projects

Tim: Could be tags, instead of (orthogonal) planes Ofer: Ideas for vectors: Timeline, adoption, ...

As an additional category, we might classify the solutions provided by the patterns into:

  • organizational and
  • social.

Test run

  • 30 Day Warranty;1/2/3/4/5

  • Common Requirements;1/4

  • Contracted Contributor;1/2/4

  • Dedicated Community Leader;1/4

  • Discover Your InnerSource;1/4/5

  • Improve Findability;? (similar to "Discover your InnerSource"?)

  • Junkyard Styled InnerSource;3/4

  • Modular Code;1/2/3/4/5

  • Review Committee;1/2

  • Change Middle Management Mindset

  • Assisted Compliance

  • Include Product Owners

  • Start as Experiment

  • Not Invented Here

  • Change Developers Mindset

  • Overcoming Project Management Time Pressures

  • Open Source Trumps InnerSource

  • Get Contributions Despite Silo Thinking

  • Contained InnerSource

Tim Yao

Another plane that would be useful is the plane of type of InnerSource. InnerSource programs can be structured to accomplish different goals: e.g.,

  • P: Product Development
  • T: Tools Development
  • I: Innovation (Proof of concepts, demos)
  • S: Shared components development
  • C: Culture

Each of these programs have some unique characteristics. E.g., sometimes in Product Development, the open source characteristic of voluntariness has to be sacrificed to ensure that sufficient development resources are dedicated to meet customer commitments and schedules. Similarly, there might be a need to limit code visibility/transparency for certain very proprietary products while making use of InnerSource methods to facilitate joint development between different business lines.

Test run

  • 30 Day Warranty;P/T/S

  • Common Requirements;P/I/S

  • Contracted Contributor;P/T/I/S

  • Dedicated Community Leader;P/T/I/C

  • Discover Your InnerSource;T/S

  • Improve Findability (s. above)

  • Junkyard Styled InnerSource;T/S/C

  • Modular Code;P/T/S

  • Review Committee;I/S/C

  • Change Middle Management Mindset

  • Assisted Compliance

  • Include Product Owners

  • Start as Experiment

  • Not Invented Here

  • Change Developers Mindset

  • Overcoming Project Management Time Pressures

  • Open Source Trumps InnerSource

  • Get Contributions Despite Silo Thinking

  • Contained InnerSource

Pattern Classification vs. Pattern Language

One lesson from PLoP 2017 was that the GoF book presented not a pattern language but a (useful) collection of patterns. Ideally, while we may have different classification systems for our InnerSource patterns, I think we want to develop a Pattern Language--a group of patterns that work together to solve a larger problem (e.g., "How do I build a new InnerSource program appropriate for my company") vs. a collection of patterns that might not have a larger goal.

Daniel Izquierdo

Another option would be to use the principles defined by Jim Jagielski in his talk "InnerSource 101 and The Apache Way"[1] as a way to characterize patterns:

  • Culture
  • Communication
  • Transparency
  • Collaboration
  • Community
  • Meritocracy

And in addition, this would have some ortogonal techniques to work on building a proper transparency (for instance) that could go from the infrastructure to be used to monitoring the process and produce surveys, training and other actions.

Another potential characterization would be to use a similar structure as existing in the organizations. This would affect all of the departments in that organization. For instance, the 'Review Committee' pattern helps with the process of letting developers work on their own and still give control to middle management and business roles. Would it make sense to have another potential characterization based on the companies structure?

  • IT/DevTeams
  • Marketing
  • Management
  • Legal
  • Business
  • Financial

Test run

  • 30 Day Warranty

  • Common Requirements

  • Contracted Contributor

  • Dedicated Community Leader

  • Discover Your InnerSource

  • Improve Findability

  • Junkyard Styled InnerSource

  • Modular Code

  • Review Committee

  • Change Middle Management Mindset

  • Assisted Compliance

  • Include Product Owners

  • Start as Experiment

  • Not Invented Here

  • Change Developers Mindset

  • Overcoming Project Management Time Pressures

  • Open Source Trumps InnerSource

  • Get Contributions Despite Silo Thinking

  • Contained InnerSource

[1] https://es.slideshare.net/jimjag/apachecon-2017-innersource-and-the-apache-way

Russ Rutledge

I like a lot of the other planes suggestions. Wanted to add one more - the point in the lifecycle of the InnerSource project. Does this pattern apply to:

  • Pre-launch (prepration to launch) an InnerSource project?
  • Launch (initial kick-off)?
  • Initial growth?
  • Broad adoption?
  • End-of-life?

Test run

  • 30 Day Warranty

  • Common Requirements

  • Contracted Contributor

  • Dedicated Community Leader

  • Discover Your InnerSource

  • Improve Findability

  • Junkyard Styled InnerSource

  • Modular Code

  • Review Committee

  • Change Middle Management Mindset

  • Assisted Compliance

  • Include Product Owners

  • Start as Experiment

  • Not Invented Here

  • Change Developers Mindset

  • Overcoming Project Management Time Pressures

  • Open Source Trumps InnerSource

  • Get Contributions Despite Silo Thinking

  • Contained InnerSource

Ofer Hermoni

tools security Culture change (probably need to separate into multiple categories) buy in trust etc. Processes / Methodology Coding practices