I realised I haven’t written on this for a while and I was asked about it again today.
Stewart’s dot twenty rule is that a piece of software is never really mature until a dot twenty release.
This was a variant of “never use a dot zero release” which has been around the industry for a long time (i.e. always wait for X.0.1).
My first written observation on my variant on this rule was back in 2006:
This is a really stupid metric of software maturity. It is, however, disturbingly accurate.
It seems to continue to be both really stupid and disturbingly accurate. The first few point releases are still going to have rough edges and once you get to about 5 you likely have something that’s intensely usable for a good number of people, by dot 10 the more complex use cases should start to be okay and once you get to dot twenty, then you could say it’s mature.
A topic for another time is how releasing often is one thing but maintaining a release is quite another.