Dates for command line should be human readable

[dalekunce]

[kamicut]

Still figuring this out. The problem is that the server we pull the data from doesn't necessarily publish the data every minute.

This is what we tried: The filenames are incremented every time the changesets are published. If we work under the assumption that a changeset file is published every minute, then we can "count minutes" from the origin, and use those dates.

However when we did that, we found that the numbers don't match up. The server sometimes takes too long to publish, and thus that minute isn't associated with a filename.

One way to go around it is to scrape every page (check link above) and associate the dates with the filenames. This would mean there is a separate process or server providing the map between dates and filenames, continuously updating.

[mikelmaron]

Another strategy would be to overshoot, and gather a few extra replication files, and then filter to the timestamp in osm-meta-util

[mikelmaron]

@Rub21 put together a library to make this calculation https://github.com/osmlab/osmreplication

@kamicut what's the best way to integrate?