On 9 August 2014 04:58, Lester Caine <lester@lsces.co.uk> wrote:
The problem here is simply 'users can choose to use or ignore at-will' if that is done at a distribution level! I've never had a problem that some of the data's accuracy can be disputed, and I do feel that the ignoring of pre-1970 data has been overcome - the Russian data is nice to see - but if the answers returned when checking historic data on a timezone depend on what data has been loaded then we are back with the problem!
Ah, that is a good point. To avoid such a mish-mash, we could strongly encourage those maintaining distributions to use only the standard data. If people are really eager to maintain an additional "tz-extended" distribution alongside that, so be it. On 9 August 2014 16:14, John Alvord <johngrahamalvord@gmail.com> wrote:
For those interested enough, there should be a variorium database for a permanent record of all the data and arguments.
IMHO, that's probably overkill. At some point, it's incumbent upon developers to adhere to best practices, and simply storing a historical UNIX timestamp under the assumption that corresponding wall clock times will never change is naïve at best. Storing UNIX timestamps alongside tz identifiers and ISO-formatted timestamps should be enough to be able to detect any resulting divergence if it's absolutely critical. Then it's up to each developer to decide which they think is more correct for their purposes. -- Tim Parenti