Open Stack

2015-11-18 16:15 GMT+09:00 Andreas Jaeger <aj at suse.com>:
> On 2015-11-18 07:30, Akihiro Motoki wrote:
>>
>> 2015-11-18 14:34 GMT+09:00 KATO Tomoyuki <tomo at dream.daynight.jp>:
>>>>
>>>> On 2015-11-17 13:05, KATO Tomoyuki wrote:
>>>>>>
>>>>>> openstack-manuals is open again.
>>>>>>
>>>>>> We still do not publish translated manuals, I'm fixing this (needs a
>>>>>> project-config change), the rest looks fine but I appreciate if you
>>>>>> keep
>>>>>> your eyes open and speak up if I missed anything,
>>>>>
>>>>>
>>>>> Andreas,
>>>>>
>>>>> Needless to say, not your overlook.
>>>>> I think we can improve translation resource generation method in recent
>>>>> days.
>>>>> As long as Liberty, we publish only install guides,
>>>>> so we can simply generate one pot file without common-rst.
>>>>> I checked our tool chain, it looks okay to improve by updating
>>>>> doc/generatepot-rst.
>>>>> Also, openstack-doc-tools skip common-rst process if it doesn't exist.
>>>>>
>>>>> Thought?
>>>>>
>>>>> [1]
>>>>> https://git.openstack.org/cgit/openstack/openstack-manuals/tree/tools/generatepot-rst.sh#n38
>>>>> [2]
>>>>> https://git.openstack.org/cgit/openstack/openstack-doc-tools/tree/bin/doc-tools-check-languages#n63
>>>>
>>>>
>>>> We could do - but wouldn't that make it harder to copy translations back
>>>> to master?
>>>
>>>
>>> Here is a refined idea: https://review.openstack.org/#/c/246751/
>>> I'm not sure I18n team want to copy install guide translations back to
>>> master,
>>> but it uses standard common-rst.pot, so I do not think it make it harder
>>> to copy process.
>>> This reduces messages from 2228 to 1849.
>>> # Huge translation resources sometimes demotivate translators :)
>
>
> I like the idea, thanks!
>
>> Talking about translation merge back to master, IMO it should be avoided,
>> at least we should be careful.
>> In my experience, merging in Zanata sometimes leads to some confusion.
>> If we use a merge in Zanata, a target branch of a merge should be
>> marked read-only.
>> Otherwise, some works in the target branch will be lost.
>
>
> I suggest to only copy over strings that are not translated in master,
> Zanata supports this,

It should work in most cases except some corner cases.

During the merge back, rejected strings in master is reset to
translated status. I reported the detail in I18N list before.
When Liberty translation was merged back into the master branch,
I needed to re-reject such strings... it was about 100 strings and unproductive.
If a language team can choose to use the merge or not, it would be great
but at now the infra team does at one time and there is no choice for
each language team.

It is just my opinion and I am okay with either decision.
If it is useful for all other teams, I don't oppose it.

Based on my experience, I am always try to keep both versions synced
as soon as possible to avoid unexpected behaviors in Japanese translation,
so any decision should not affect us a lot.

Akihiro


>
>
> Andreas
> --
>  Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
>   SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
>    GF: Felix Imendörffer, Jane Smithard, Graham Norton,
>        HRB 21284 (AG Nürnberg)
>     GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
>