Generated with

	ql -db vino.ql 'select id(), name, rating, country, picture from wines' > dumped/wines.csv
	ql -db vino.ql 'select id(), wine, content from comments' > dumped/comments.csv

\o/

this is in preparing to move to sqlite as the backend

Other 404s are already logged in their respective handlers.

The path `/` alone is a wildcard, which also matches things like
`/foo`, `/potatoe`, `/yes/this/works/` and such. We don't want that.
Send an HTTP 404 for any path not matched otherwise and not equal to
`/`.

Login is done by clicking the user button in the top right.

`hash` is not really specific.

It's easier to use and set up than logrus, and the API is a bit nicer
(no distinction between a logger and a log entry, for example).

This way, failed authentication also has a request ID and such

If we're behing a proxy such as Caddy, the remote address will be that
of the proxy, so we have to use the X-Forwarded-For header that the
proxy injects instead.

This makes it easier to re-use the session after a while, and to
implement a logout option (by just deleting the session)

It fits the need more naturally.

This makes it easier to deploy the whole thing on FreeBSD

This makes it a bit easier to track things like permissions.

Makes it impossible to see which users have the same password.