[Myrmornis]

> Custom data structures can be composed from the well understood builtins, rolled in under 100 lines of code and can can exist close to the place they are used (yes repeated!). The effect of this approach on readability, maintainability, decoupling, ... adds so much more value to the whole lifecycle, than the cost of the omission.

It's interesting to me that this philosophy comes from the Go designers at Google, and that Google is also well known for keeping vast amounts of source code advancing in lock-step in a single repository. From reading the recent article on Google's source code repository structure, I believe that being able to reuse code (e.g. data structure implementations) without versioning headaches is one of the intended and actual benefits. It's of course not that surprising that two different areas (Go design and repository structure) might pull in two different directions, but these are two important high level issues so it does seem a little inconsistent to me.