MongoDB has finally relented and changed the default behavior for handling errors when writing in the new version of their client. Previously the default behavior was not to wait and see if the write really worked. Unless you explicitly asked for it, a failed write operation might go unnoticed.

MongoDB has gotten a lot of flak for the previous behavior and it has been a big reason for some people to abandon MongoDB in favor of other database solutions. I have never been able to work up much sympathy for users that has been complaining about having lost important data because of not setting the write behavior correctly for their application. Using a database in production without having at least a basic understanding about how it works is a sure way of inviting trouble regardless of which database you use.

But on the other hand I think it is a very good move of MongoDB that they should have done long ago. Having writes fail silently by default is a bad idea that is really counterintuitive for most people. (The reason for the old behavior is clearer after having read the previously mentioned blog post, although the reason for the delay in making this change is not).

I believe a lesson to be learned here is that if you are designing such a system as for instance a database, you should really think twice before choosing speed over safety as default behavior. Most users probably want safety and predictable behavior as default and increased speed as an option, not the other way around. Especially when it is as easy as changing one argument to choose between them..