I wasn't referring to concurrency within package scope, but data crossing package boundaries (the earlier post says libraries' are responsible to ensure data passed *to* them is safe). The map type reveals an implementation detail by documenting that it isn't safe for parallel use.
On Friday, June 24, 2016 at 8:52:40 PM UTC-7, Henry wrote: > > The implementation of map is not a valid argument to "unboxing the black > boxes". > > The question to ask is who does the concurrency. In the case of map, if > the user is the one accessing the map concurrently, then it is the user's > responsibility to ensure the map is concurrent safe. If the map internally > implements some concurrent processing, then the map has the responsibility > to ensure data integrity during the concurrent processing. The map's > implementation should be invisible to the user. > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.