Re: [PING^3] RFC [PATCH] BZ#1077902: New API gettimezone

[Paul Eggert <eggert at cs dot ucla dot edu>]

P J P wrote:
> > On Thursday, 29 May 2014 7:25 PM, Paul Eggert wrote:
> > Yes, application dependent.  Scheduling applications, for example, are
> > more interested in times in the near future, whereas date-of-birth
> > applications are more interested in times in the medium-distant past.
> > They have different needs for the "best fit".
>
>    I doubt if such applications have any need for a 'best fit' definition, because they do have access to the system's tzdata files.

No, these applications typically don't access the system's tzdata files 
directly.  They use the glibc API (or its POSIX subset) to get timezone 
information.

> > If the problem cannot easily be solved in glibc then let's be honest and
> > leave things alone, and let application developers continue to solve the
> > problem in the same simple and portable way that they've solved it for
> > decades.
>
>    That is precisely what the new API for which the patch was submitted early in this thread aims to do.

No, the patch proposes a change to the API, which would require programs 
to change if they want to use the new facility.  This is not leaving 
things alone.

> It'll greatly help to discuss whether 'best fit' is the TZ string at the end of the tzdata file, when it is present, or any other line from the tzdata file.

It's not the 'best fit'.  There is no such thing as a global 'best fit'; 
as I explained to you, and as Russ Allbery explained in 
<https://sourceware.org/ml/libc-alpha/2014-05/msg00023.html>.

At this point I feel like we are going around in circles, as you are not 
responding to the issues raised and are merely repeating your proposal.