Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:117492
Date: Thu, 7 Apr 2022 11:33:36 +0100 (BST)
To: Nikita Popov <nikita.ppv@gmail.com>
cc: PHP Developers Mailing List <internals@lists.php.net>
In-Reply-To: <CAF+90c-bTC8om=_c61kFAhcqe+NzwMv6JX7eyoXLY7DJHJm8Zw@mail.gmail.com>
Message-ID: <alpine.DEB.2.23.453.2204071119150.2094339@singlemalt.home.derickrethans.nl>
References: <alpine.DEB.2.23.453.2204071026180.2094339@singlemalt.home.derickrethans.nl> <CAF+90c-bTC8om=_c61kFAhcqe+NzwMv6JX7eyoXLY7DJHJm8Zw@mail.gmail.com>
User-Agent: Alpine 2.23 (DEB 453 2020-06-18)
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Subject: Re: [PHP-DEV] Timezone Rules, which dataset to pick?
From: derick@php.net (Derick Rethans)

On Thu, 7 Apr 2022, Nikita Popov wrote:

> On Thu, Apr 7, 2022 at 11:34 AM Derick Rethans <derick@php.net> wrote:
> 
> > Java's date/time maintainer has created a fork based on the original 
> > Iana TZ data to put back some of the removed/deprecated data to 
> > better serve their users, and I would think that this is also best 
> > suited as a data set for PHP.
> >
> > If you want to read about the intricacies, see: 
> > https://github.com/JodaOrg/global-tz#rationale
> >
> > But this does mean a divergence from the "official" TZ data, 
> > although Joda's data is arguably better. My recommendation is that 
> > from the 2022b release we switch to Joda's version. (I will today 
> > merge in the 2022a data from the Iana source.)
> 
> Keeping in mind that people deploying PHP on Linux usually end up 
> using OS-provided zoneinfo, do you know which source distros base that 
> on? I think we should follow the distro-consensus here, whatever that 
> may be.

NetBSD has switched, FreeBSD is entertaining the idea and offers the new 
data as an option. I haven't seen any comments from Linux distributions.

Our long standing stance on distributions patching PHP for their own use 
has always been "you're on your own". The data these users receive is 
already degraded, due to their patching, and that Iana TZ has already 
removed some historical data since a few releases. As this removal only 
pertains pre-1970 data, where the Joda time version provides accurate 
data for all countries (and not just aribitrary ones), I both don't see 
that a divergence is a big problem, nor do I find it an argument for 
releasing known-wrong data.

It is mostly African countries now with known-wrong data, but the TZ 
maintainer tried merging Amsterdam and Brussels (removing likely-correct 
pre-1970 Dutch data in favour of certainly-wrong pre-1970 Belgium data), 
and several other European countries. That was stopped for now, but I am 
certainly he'll try again.

I don't think it is acceptable for us to say that we're okay with some 
countries data to be known-wrong.

cheers,
Derick