Re: [PATCH] Re: date timezone question

[Paul Eggert]

address@hidden writes:

> http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap08.html#tag_08_03
> `info libc "TZ Variable"`
> suggests that implementation dependent TZs should start with a colon.

That advice is admirable but is quite rarely followed.  In practice,
people almost invariably use something like TZ="Asia/Tokyo".

The ":" is there because POSIX wanted to provide a standard way
of saying "this is a local extension", a way that POSIX itself
would never take away in any future standard.  But the tz code
also provides the more commonly-used notation TZ="Asia/Tokyo";
this is a pure extension to POSIX, so it's conforming behavior,
but (unlike ":") a future edition of POSIX theoretically could
assign a different meaning to TZ="Asia/Tokyo".

Of course in practice this will never happen due to the common
practice we're talking about, so in practice the Olson names are safe
without the ":", and that's what most people use.

> So I propose to only support that in date. for e.g.
>
> date --date "09:00 :America/Los_Angeles"

But what about POSIX-conforming TZ strings?  Shouldn't we support them
too?  E.g., TZ="PST8" works as a TZ setting, so there should be some
notation for it in the date string.  A more complicated example is
TZ="CET-1CEST,M3.5.0,M10.5.0/3", which should also be supported.
Basically, the full POSIX TZ syntax should be supported, angle
brackets and all.