I've followed the discussions about historical data with interest. I would like to keep historical data in the tzdb, and I understand why I'm in the minority. But I wonder if there couldn't be a compromise? We've implemented a .NET library that uses the tzdb by interpreting[1] the database rather than by compiling it. So for us, we only deploy the tzdb text files that the customer cares about. Some customers are strictly North American concerns, so we only put the northamerica file in their configuration; others need the whole database of 500-some zones. I think other posters have suggested separating pre-1970 data out into a separate file or file set. This would be perfect for our implementation. For the zic implementation, it seems like compilation is an all-or-nothing deal right now. But would it be difficult to create a build script that includes the historical data to the zic package if someone needs it, in addition to the post-1970 build script? Sorry if the question seems naïve. I just think there's room to compromise here. David Braverman Inner Drive Technology Chicago [1] The tzdb files actually get compiled down to an intermediate set of objects at application start, then live on in a singleton. The process takes about a second on an Azure Extra-Small instance. More info here: https://www.inner-drive.com/Demo/Timezones.aspx
David Braverman wrote:
would it be difficult to create a build script that includes the historical data to the zic package if someone needs it, in addition to the post-1970 build script?
No, that should be doable. Nobody's done it yet, though. (Zefram has a prototype, which still has some kinks.) We will probably do something like that at some point, if only to support the few old entries we have lying around. There's a more important issue than the technical mechanism. Even to cover just standard 24-hour time with the Gregorian calendar, a large amount of information would be needed to fill out the pre-1970 data: it'd easily be an order of magnitude larger than what we have now. Who'd do the work to maintain this? "Maintain" doesn't simply mean "make sure it gets through zic without complaint": one must also make sure it's well-sourced and that there aren't any copyright violations in it, and deal with any conflicts among sources, including the inevitable political conflicts among contributors. As this is out of our traditional scope, and we're already swamped by too much email, it may be better to start a new mailing list just for the topic of pre-1970 zones, a mailing list that could discuss creating the historical database. We'd gladly refer people to such a list.
participants (2)
-
David Braverman -
Paul Eggert