The current ccache code is split in backends with the idea that each one needed very different manipulations, but we should try as much as possible to use just krb5 functions, so that any future ccache enabled mechanism is automatically supported.

The main reason we have backends is that the FILE cache is somewhat special in that normally we use a randomized ccache name and we need to keep this name around in order to be able to renew or reinit while maintaining the current name as used by active user sessions.

However in case random names to identify the whole ccache are not used like is the norm for cache collections, we should use exclusively krb5 functions to initialize and destroy the ccaches. With the notable exception that we may need to create the containing directory for DIR caches.

By using only krb5 functions we do not need to have special cases for creating and destroying anmd checking ccaches.

For the FILE case we could also abstract away handling of randomized name caches fro the general code, so that in the end the only real special operations are checking for existing ccaches in order to create the containing directory and checking for recorded names only for randomized ones.