Fog Creek Software
Discussion Board




Other management ideas from the open-source world?

Today, Joel talks about making frequent releases to a select group of customers.  Anyone who's ever visited the open source world knows this idea is a form of "unstable" or incremental versions, that occurs spontaneously in the opensource ecosystem.

Are there any more project management ideas that people use, that were inspired by open source development?  For me, I've taken a lot of ideas from distributed dev, using central project info areas and making reasonably few assumptions about peoples' development environments.

Hannes Glinka
Monday, April 29, 2002

The idea of having huge mailing lists on which you chat about things in the project. It's a low-intensity way of spreading information about.

Most industrial projects get run by havign big formal documents that a) lag behind the actual project and b) take a lot of process to run and c) only document a state, not an evolution.

By contrast the open-source model of mailing list archives gives you a log of why decisions were taken as well as what decisions were taken - the conversations about what didn't work are on record.

Katie Lucas
Monday, April 29, 2002

log = yet another document, not "formal" if you like, but hey, what's in a word.

By its very nature open source documentation is unweildy.
End result, a big ball of mud. Try building the equivalent of an open source skyscraper, disaster.

Tony
Monday, April 29, 2002

Possibly true, but then the "formal" and "proper" school of software development produces... what was that survey result? An average of 30% of projects cancelled due to massive budget over-runs, and something like only 16% are considered successes.

Whenever you talk about introducing new techniques, the argument is always "That might not work!!" which conveniently disregards the past information that doing it the old way is known to more likely than not fail.

Katie Lucas
Tuesday, April 30, 2002

Yes I see your point Katie, but would Airbus ever allow an open source methodology for flight software? If they did would anybody ever get on an Airbus again?

Tony
Tuesday, April 30, 2002

FWIW, I am only concerned about adapting certain best practices and changing them to suit the situation.  Airbus should not deliver software like the Linux kernel team, but it is very likely they have something to learn from others. 

For a while, I studied GE carefully, especially the Not Invented Here syndrome.  Their solution was not just to steal others' ideas, but to reshape and fit to their own environment.  That included throwing out new ideas whole but being inspired by those failures to find a brighter solution.

Hannes Glinka
Tuesday, April 30, 2002

Yeah, but airbus don't have something like a 2/3 failure rate to start with. If they did, they might be wondering why people don't get on the planes. We have that level of failure rate, but our management isn't asking whether we should change our methods, they just think they haven't bullied the developers enough yet. It's like airbus wandering down and screaming at the welders to weld harder and fill in more paperwork about their welding because no-one can understand why planes with only one wing don't get off the ground and crash at the end of the runway... and suggesting adding an extra wing is heresy - they've ALWAYS made planes that way, no-one's going to change anytime soon!

{The engineering world, by the way, is SCARY for some of its results. My dad's just retired from the aerospace industry. In 36 years of working for the same company, they have NEVER had a unit fail in service...}

Katie Lucas
Tuesday, April 30, 2002

Actually I wonder if airbus flight software is opensource or not?

Daniel Shchyokin
Wednesday, May 01, 2002

Katie, I like your analogies for management of software development proejcts. (Yelling at the welders to work faster...)

So much of the criticism levelled at software development, such as failing to meet deadlines, cost overruns, and so on, are really failures by management, not development.

For example, failing to meet deadlines is better defined as a failure by management to understand workload and requirements, and thus to set accurate deadlines.

Hugh Wells
Wednesday, May 01, 2002

daniel, no, flight control software is not open source. It is subject to incredibly exhaustive checking, with rigorous code reviews and compliance testing. No changes to designs are allowed without formal evaluation by all stake holders and assessment of downstream consequences.

In operation, flight control consists of multiple modules that continually check each other. On some aircraft, the modules run on different processors as well.

Hugh Well
Wednesday, May 01, 2002

For a software technology to be "open source", its source code is available openly. If a major company made its operating system available then its open source. From what I recally, the term "open source' does not denote a project management paradigm.

Indeed, if airbus did make some of its software open source would that mean that any and all sound project managent pardigms would be cast aside.

Certainly I agree most open source ::projects:: could probably use better management and certainly better documentation. I wanted to emphasize that the term "open source" pretty much means that the source code is made generally available rather than being kept secret.

A. Coder
Tuesday, July 22, 2003

*  Recent Topics

*  Fog Creek Home