Implementation and benchmarks of the two methods of generating unique numbers described in this blog post: http://nf.id.au/concurrency-patterns-a-source-of-unique-numbe

The output of 'gomake bench':

uniq.BenchmarkMutex 10000000 173 ns/op uniq.BenchmarkChannel 10000000 224 ns/op

The channel-based approach takes 1.29x the time of the mutex-based approach.

Depending on the specific application, and your opinion on the aesthetics of the two approaches, this can be an acceptable tradeoff (it is, IMO).

Andrew Gerrand [email protected]