I dunno. I think it would be a lot less code, and a lot faster, (in the final sketch) if you wrote your own "yesterday" function that knew about days in each month and leap years, and so on, compared with including a library based on the unix "seconds since 1970" stuff. I mean, unix time is elegant and fits nicely in filesystem data structures and stuff, but it generally takes a lot of multiplications and divisions to go between the "seconds" and the current month/day/year/etc.