I don't consider redis durable for similar reasons as mongo - yet redis doesn't hide what it is or what it lacks in terms of durability... so it doesn't bother me that it isn't.
but mongo often downplays it's lack of durability, which makes me wary of using it or recommending it because people don't know how to use it.
This is exactly the problem with MongoDB. It's practical use is a semi-durable, in-memory data store, but they advertise it as a data store of record that can be used for large, >RAM sets of any conceivable use case.
It seemed a bit dishonest seeing benchmarks that showed how MongoDB beats CouchDB in write-performance tests when there was no acknowledgement of write requests comming back to the client.
Not saying that MongoDB wouldn't be faster with full commits turn on, it is just that their design decision to pick that as the default and calling their product a 'database' seemed very strange at best, and dishonest at worst.
For the record: we (MongoDB) have a policy of not posting official Benchmarks. Could you clarify what benchmarks you are referring to that were posted and "dishonest"?
The only benchmarks we have ever done have been for internal testing comparisons between different versions.
Previously there were several third party benchmarks on this page, which we recently took down to stand firmer on our benchmark policy. In the interest of full disclosure however, you may examine the page history to verify that we have always clarified these as unaffiliated third party sourced.
but mongo often downplays it's lack of durability, which makes me wary of using it or recommending it because people don't know how to use it.