Alistair - I'm having trouble grokking CrystalClear as medium discipline with its 20 required deliverables. Please say more. --r
says, "you really ought to do these things. If you don't do them, you are in violation." That is fairly low-tolerance. Tolerance
says, "there are lots of ways of doing these things. Choose one that works for you." Size
says, "there are these many things to produce or do." XP is by far the smallest methodology I have ever seen. I can name all its practices and deliverables on one slide in Powerpoint. That is no less than astonishing. 20 deliverables, including requirements, project planning, design, code, test and migration, is also tiny, compared to most methodologies. Just not as tiny as XP. But number of deliverables is not a reflection of the discipline or tolerance of the methodology.
XP says, if you don't do PairProgramming
, you are not doing XP. If you don't have UnitTest
s, you are not doing XP. etc. So for its small size, it is not tolerant.
is saying, you have to have these things in your requirements, but you can write them in any of a set of ways. You have to have an incremental delivery plan, but you can run your increments in any of the substitutable ways. And you have to have very close communications and process checks (as in the VcapsProject
) for it all to work. Whether the software comes out or not then depends on the quality of the people, the quality of the communications, and the quality of other things, like the executive sponsor. Compare with the critical success factors I named in the SurvivingObjectOrientedProjects
- willingness to find and fix failing ideas,
- incremental staging,
- executive sponsor, project manager, technical leader,
- habit of delivering
The 20 deliverables were those I could find when I visited successful projects, plus one I added gratuitously (EmbellishmentIsThePitfallOfTheMethodologist
) - regression tests. Most successful projects I have visited did not have regression tests, and they shipped anyway. However, those that did have regression tests shipped more easily. So I freely added regression tests.
p.s. a critical success factor I determined years ago regarding methodology design is - if the people don't want to use it, they won't. So to be successful, it must first be used. To be used, it must be both small and palatable. The gamble on CrystalClearMethodology
is to find something small and palatable enough, and still good enough to allow the software come out the other end. -- Alistair