Re: Re: A patch added to mozc by submit-request 114205

On Wed, Apr 18, 2012 at 4:14 PM, Takashi Iwai <[email protected]> wrote:
> At Wed, 18 Apr 2012 15:03:04 +0800,
> Weng Xuetian wrote:
>>
>> On Wed, Apr 18, 2012 at 2:30 PM, Marguerite Su <[email protected]> wrote:
>> > 2012/4/18 Fuminobu TAKEYAMA <[email protected]>:
>> >> Hi, Marguerite Su and M17N maintainers,
>> >>
>> >> I am confused by the big patch for Fcitx added to the mozc package yesterday
>> >> without any reviews and discussion.
>> >>
>> >> https://build.opensuse.org/request/show/114205
>> >>
>> >>
>> >
>> > Hi, ftake,
>> >
>> > yes, I don't why either, when I created the request, no reviewer is
>> > automatically added.
>> >
>> > only me, and I have no review option but to accept.
>> >
>> >> My concern is that this patch decrease maintainability of that package.
>> >>
>> >> As you know, Mozc's release cycle is very short.
>> >> Who will maintain the patch?
>> >> Will Fcitx maintainers (including the patch developer?) release new patch
>> >> for every Mozc release?
>> >> I cannot think they will do so, as long as I read their website.
>> >
>> > Fcitx developers and maintainers will do.
>> >
>> > I think you just see http://code.google.com/p/fcitx/download. actually
>> > it's their "release" website. fcitx-mozc is fast developed at
>> > https://github.com/fcitx/fcitx-mozc
>> >
>> >>
>> >> This is why the openSUSE packaging policy says "Ask upstream first", right?
>> >>
>> >>
>> >
>> > absolutely right, but this time it's Fcitx upstream developers that
>> > asked me to add so.
>> >
>> >
>> >> It is, however, good news that fcitx support Mozc.
>> >> I guess this can be demonstrated by creating separated OBS package like
>> >> fcitx-mozc if some header files of mozc are packaged.
>> >>
>> >
>> > yes, I'm glad to say Mozc users have a third input method to taste.
>> >
>> > it's not possible I have to say. you know how Mozc is developed and
>> > packaged. there's actually no such header file. fcitx in M17N has many
>> > derivatives, and it has a not-released-yet fcitx-anthy in
>> > home:opensuse_zh. if it could be done that way, it would already be
>> > done. I bcc-ed fcitx core developer, and he'll explain that tech thing
>> > in detail.
>> >
>> >>
>> >> Anyway, I would like to roll back for now because mozc is now really essential
>> >> package for Japanese users.
>> >> If the patch is the best solution and users need fcitx support,
>> >> shall we add the patch again?
>> >>
>> >
>> > Fcitx-mozc is also essential for some group of fcitx Japanese users. I
>> > think as packagers we have no obvious reason to drop a successful and
>> > well-maintained feature. by adding it, fcitx OBS maintainers will also
>> > have to maintain mozc, there're 4 active maintainers on OBS,
>> > personally I think it's good to add maintianers to such a essential
>> > package.
>> >
>> > openSUSE is the best distro that fcitx supports. fcitx developers are
>> > also on OBS( home:csslayer:fcitx* ). so there's no need to worry the
>> > unmaintained thing that early. it really hurts someone's feeling.
>> >
>> >> --
>> >> Fuminobu TAKEYAMA
>>
>> Hi,
>> I'm one of the fcitx developer.
>>
>> For your question:
>> Yes, fcitx will release new patch against every new mozc release, it's
>> supported by fcitx upstream.
>>
>> There is several reason that I can't and don't want to make fcitx-mozc upstream.
>> 1. mozc is not open developed, though it's open source.
>> 2. different release schedule might be conflict with each other, the
>> best solution from my point of view is develop outside mozc.
>> 3. mozc will not become a standalone library in near future (Though I
>> tried to ask them to do so).
>>
>> For the patch, it only adds new file, and will not break any existing things.
>>
>> For other distribution, debian and fedora, also include uim-mozc,
>> which is not mozc upstream too. Add fcitx-mozc for debian is also on
>> the way. They also use the patch in order to get uim compiled, that's
>> another reason that I choose to release patch.
>>
>> For potential users, I don't want to talk about this too much, but
>> there are request on Fcitx issue list, for adding Japanese support,
>> and anthy development it is said that have moved to mozc, and mozc it
>> much more easy to develop (Even must use some special buildsystem).
>
> OK, so judging from you and Marguerite's comments, we can take the
> patch as is for now.  But I suggest to makes easier to strip it off,
> e.g. via %if %with_fcitx or such.  Could you resubmit the package in
> that way?
>

I'll try. (because I'm not familiar with conditional builds...

> Also, at the next time, please announce or ask on m17n (or
> opensuse-factory) ML before doing non-trivial updates like this.
> Otherwise it'll surprise people unexpectedly.
>
> Basically osc submitreq review is the last step, not the first nor the
> sole step for integration.  If a new feature is added or a feature is
> changed, it should be announced to users publicly on ML beforehand.

it's my fault. I don't even know there's a m17n ML before today.

>
>
> thanks,
>
> Takashi
-- 
To unsubscribe, e-mail: [email protected]
To contact the owner, e-mail: [email protected]