I like milestone based release models.
The advantages I find in this model are in particular beneficial for MySQL. What I find good about this model are:
- Things are unstable for shorter periods. Even if some feature is not full stable in some milestone, the model encourages that such a feature is fixed on higher priority.
- It is easy to create a priority ranking for new features. Moreover, priorities are expressed more by chronological time of development, less by "how many people are working on it".
- The model pushes towards rapid development, since you can't release M5 before M4 is complete.
The last versions of MySQL took long time to complete. Take 5.1, for example: partitioning and event scheduling were long considered GA before row-based replication was half stable. Consider the so small but useful sub-second slow logs; the variables made dynamic in 5.1 (slow log again, for example); the new INFORMATION_SCHEMA tables.
Most of these took very little time to develop and stabilize, but had to wait extra couple of years till the entire version was evetually released.
There is a matter of stability question. Is M3 stable? Is M2?
Smaller, successive changes make for stabilized milestones; but this is vague: 5.5 is based on 5.4, in itself BETA. So is any 5.5 milestone stable?
Such issues need to be resolved. Like everything else in life, this model can be abused to no good, or can be properly implemented to make good success.
I'm glad to see google patches making it into the mainstream; I would love to see more community work merging in. I would love to have the community have a say about what should go into the next milestones.