<<On Wed, 31 Jan 2001 12:34:07 -0500, "Seeds, Glen" <Glen.Seeds@Cognos.COM> said:
First, the daylight saving offset is not specified. Yes, we all "know" it's an hour (to be ADDED, in this case), but a standard needs to specify this.
Actually, we don't know that. Several places have observed one-half or two-hour shifts in the past, and there is no guarantee that the politicians will not come up with new ones in the future. An examination of the timezone database shows eighteen different locations where this has been done, of which one (Lord Howe Island) still uses a non-unity-hour summer-time adjustment.
It turns out that you can achieve all of this with mktime() and gmtime(). Given that, perhaps a better strategy would be to guide applications toward that, and away from tzset().
I agree, and I think the denizens of the timezone mailing-list (cc'ed) would probably agree as well. (Speaking as an implementor for a BSD-derived system that has never implemented and probably will never implement the `tzoffset' and `daylight' variables.) -GAWollman