Moving the Japanese Era data out of the locales

Moving the Japanese Era data out of the locales

[Florian Weimer]

We have some level of Japanese Era support in strftime:

$ LC_ALL="ja_JP.utf-8" date +%Ec
平成30年03月14日 10時41分04秒
$ LC_ALL="ja_JP.utf-8" date +%EY
平成30年

(date may actually use a separate copy of the strftime code, but I 
believe it is substantially identical to the glibc implementation.)

There is going to be a transition in 2019 (bug 22964).  Having this in 
locale information is a bit problematic for (some) distributions because 
they have to update glibc for all users, on all release branches, just 
to deliver this change.

Could we synthesize this data from a dedicated file in 
/usr/share/locale, by changing the code in time/era.c?  Then 
distributions can ship this file as part of tzdata, which has a 
different update procedure from glibc.

Thanks,
Florian

Re: Moving the Japanese Era data out of the locales

[Carlos O'Donell]

On 03/14/2018 04:44 AM, Florian Weimer wrote:
> We have some level of Japanese Era support in strftime:
> 
> $ LC_ALL="ja_JP.utf-8" date +%Ec 平成30年03月14日 10時41分04秒 $
> LC_ALL="ja_JP.utf-8" date +%EY 平成30年
> 
> (date may actually use a separate copy of the strftime code, but I
> believe it is substantially identical to the glibc implementation.)
> 
> There is going to be a transition in 2019 (bug 22964).  Having this
> in locale information is a bit problematic for (some) distributions
> because they have to update glibc for all users, on all release
> branches, just to deliver this change.
>
> Could we synthesize this data from a dedicated file in
> /usr/share/locale, by changing the code in time/era.c?  Then
> distributions can ship this file as part of tzdata, which has a
> different update procedure from glibc.
We would have to add net new code to handle this shared resource,
all for what is really a distribution packaging issue.

IMO it would be of higher value to fix the downstream distributions
to ship locales in distinct packages to allow them to be updated
independently of glibc.

This way the problem is solved the next time we have an era change
in 10-80 years.

Cheers,
Carlos.