Open Source as a different Silver Bullet Carlo Daffara Conecta Srl, AMOS project cdaffara@conecta.it...

Preview:

Citation preview

Open Source as a different Silver Bullet

Carlo Daffara

Conecta Srl, AMOS projectcdaffara@conecta.it

http://www.amosproject.org

Advancing the research Agenda on Free/Open source Software

Brussels, 14 Oct. 2002

• Software is pervasive, our preferred way to add intelligence to objects

• Software is the basis for most of the current world economic infrastructure, and is behind the improvements in health care, logistics, manufacturing...

• Of 24 technologies of the future, 18 are software intensive

• Despite this, we are still unable to cope with software complexity

• We are in the same “software crisis” first described in 1968

• The Standish Group reports that, for US companies in 1998, regarding software/IT projects:– 26% succeeded (it was 27% in 1996)– 28% failed– 46% are “challenged” (over budget, over time,

with fewer functionality). We must consider that many CIO now prefer to turn a failure in a “challenge” to avoid a negative loss to be written off immediately on the balance sheets

• Why? Brooks described some structural reasons in his “No Silver Bullet” paper (1987), and formulated his law as: “there is no single development, in either technology or management technique, that by itself promises even one order of magnitude improvement in productivity, reliability, in simplicity”

• Despite the years, and several trials, the basic assumptions still hold true

• So, it seems that there is no hope...

• Brooks describes some potential “attacks” that collectively may improve the situation, and mentions:– Buy, don't build– Incremental development, prototyping, testing– Great designers

• In the open source world, we would say:– Reuse code, don't build– Incremental development, prototyping, testing– Great designers (open source projects have

usually a “core” group that plan, design, develops)

• If that's true, the open source software development process shuld show greater speed/quality

• Some preliminary results seems to say that:– Comparing productivity in loc/hour, open

source vs. commercial: 1 to 1.5, despite open source programmers average half the hours

– Comparing quality: a test of Apache evolution shows defect density significantly lower [Mockus, Fielding, Herbsleb]

• But we need to prove it!• We need research, comparable to that made by

Standish, to see if open source (both as methodology and as potential resource to be tapped) improves the final outcomes

• We need to discover what hurdles and barriers exist in the adoption of open source, and see if those can be overcomed through government (including university) or private support

• We must shamelessly use what has already been developed, and try to leverage different projects to create synergies

Questions?

Recommended