> And likewise, that it's okay for us libraries to develop features which
> are used only by significant minorities of our users (important to remember
> what our logs show is really significant minorities of _uses_. All users
> using a feature 1% of the time can show up the same as 1% of users using a
> feature 100% of the time). We are not lowest common denominator, while we
> need to make our interfaces _usable_ by everyone (lowest common denominator
> perhaps), it's part of our mission to provide functionality in those
> interfaces for especially sophisticated uses that won't be used by everyone
> all the time.
Exactly. When designing services, it is natural for people to use
popularity based metrics (i.e. how much does something get used) as an
indicator of quality. But such logic is dangerous, because if we accept it
we must agree that McDonalds makes the best food on earth.
The value is of what we produce is ultimately in the impact, and
limitations in the tools commonly used to quantify this should not be used
as a substitute for thinking.