IBM translations to contribute
Hi i18n Friends, I'm happy to share that we are proceeding with our plan to share our IBM translations with the OpenStack community. To review and follow up on what was shared in the 2015-08-20 i18n team meeting [1]: - We would like to contribute our IBM translations for projects ceilometer, glance, heat, nova, ironic, neutron, cinder, keystone, and swift. Note that Horizon is not in this list because the community has been focused on translating Horizon; I'm concerned there may be excessive terminology conflicts. - We have translations for de es fr it ja ko_KR pt_BR ru zh_CN zh_TW - These contributions are based on our Kilo translations. They have been reviewed and tested. - In order to facilitate an informal review by the translation teams before uploading I've made the translations available at https://github.com/doug-fish/openstack-translations . The repository is private. If you'd like access, just share your github id with me via email (drfish@us.ibm.com) or IRC (doug-fish) and I'll give you access. - Segments that have been updated with IBM translations have a comment "Contributed by IBM" added. This is to enable reviewing; I don't expect these comments to remain in Zanata after the upload. At a high level, our process is that we are extracting the existing PO files from Zanata, then using Babel based code to compare our IBM translations with the community ones and filling in any blanks. We've taken care not to overwrite any existing translations. These proposed files are based on what we extracted from Zanata recently (today). Our intent is to run this tooling again after the review period to pick up any community translations that have occurred. I'm aware that translations for Nova are not updated in Zanata yet, but again, I expect to handle this by re-running our tools. I'm working with Thomas Cocozzello (tjcocozz) and Lucas Palm (lapalm) who have coded our tools and scripts to handle this. They are requesting to join the language teams in Zanata so that if there are no concerns noted during the review they can upload translation contributions. I plan to ask Tom and Lucas to upload these files on 2015-09-14 unless there are concerns noted during the reviews. Please let me know if you have questions or concerns! Doug 1. http://eavesdrop.openstack.org/meetings/openstack_i18n/2015/openstack_i18n.2... Doug Fish
On Wed, Sep 9, 2015, at 02:34 PM, Douglas Fish wrote:
After reading the meeting log it appears that using a git repo of some sort to perform easy diffing was the suggested way to go through this process. My only concern with that is these translations are being treated in a special manner to the detriment of other contributors that have to go through the normal translation process in Zanata (and previously Transifex). At the very least current translators may not realize that there is a bunch of work done because it is currently hidden behind a private Github repo. My suggestion would be to push these translations into Zanata the same way any other translator would. That communicates to other translators where to focus both review and translation efforts. And if we need better diffing abilities that may make for a good feature request to Zanata? (note this is based on my understanding that translations in Zanata have to be approved/accepted and I don't think that we should bypass that process as all other translators have to go through it). If that is not feasible for practical reasons my suggestion would be to use Gerrit instead of Github and perform the review in the open. This way we have record of the reviews and anyone can participate using the tools already in use for reviewing git changes by OpenStack. You should be able to push changes to various projects updating their .po(t) files in order to get the diffs. If you do it atop the current translation change proposals in Gerrit you should get the correct diffs out of it. I know I would personally be a lot more comfortable with this if Andreas' could weigh in on it. Maybe we can get his feedback before making too many changes? Also, do let me know if the infrastructure team can do anything to help IBM contribute upstream first so that we can avoid needing to work through special cases like this in the future. If there are deficiencies in the system/tooling it would be great to fix them. Clark
Clark Boylan <cboylan@sapwetik.org> wrote on 09/09/2015 08:26:51 PM: period
Clark, I think you and I have a lot of agreement about the translation process. I agree it would be better if contributions like this could be handled using the normal processes. I had the same expectation you did that translations had to be approved/accepted before they were included in the service's git repositories; I now understand this is not the case. While it's true that Zanata includes a review feature, this is a post-acceptance review. Unreviewed translations are included in the exported PO files. Translation reviews happen at a later time. As a consumer of translations, it's important that they go through a process that parallels the process of the code: translations need to have a review before they are included in the normally exported PO files, just like code needs to have a review before it gets included in the git repositories. I understand this will not be an easy change. Performing reviews will take time, and our translators are already pressed for time the last few weeks of the release. We'll need to explore solutions to this problem. We need to make sure we can get broader participation in the translation process, and perhaps translations should not be made available until the first stable fixpack in order to give translators time to complete this work. Please note that we are not bypassing any features of Zanata. I've added this informal review in github based on discussion that occurred in the i18n meeting. If the translation teams found that there was a terminology or other problem with the translations we intend to contribute, there is no easy way for them to reject them all in Zanata. At the end of this informal review, these translations will be uploaded into Zanata just like any translation team member could, and the usual processes will all still apply. Andreas, do you have any concern or further suggestion? Doug
to pick up any community translations that have occurred. I'm aware
Hi, all I have talked with Clark, Doug, and some translators today. As a summary, here are my understanding to the current situation. 1. Translators welcome IBM contributions because that could help to lessen the translation efforts. 2. IBM contributions have to go through the formal translation process, including: translating in Zanata (although offline and then upload), being reviewed by reviewers, being pulled from Zanata, being reviewed in Gerrit, and passing the testing, before it gets merged finally. 3. Zanata doesn't support to pull reviewed and approved translations. All translations except rejected translations, even not reviewed translations, will be pulled. 4. Because IBM contributions are a batch of translations, translators might not have enough time to review one by one. Some translation team may want to do a pre-review before the po files are uploaded to Zanata. The key point is how to do the pre-review. Based on all the feedback I collected today, there are several options. Option #1: Pre-review in a private github repo. This one is not good because it's private, not public. Option #2: Pre-review in Gerrit The good thing is that it's public and under Openstack governance. The bad thing is that it'll take draft translations to developers' sight. Developers may get confused when they see the translation patch. What's more, messages in IBM's po files might in a different order, which makes the comparing with existing po files difficult. Option #3: Pre-review in Zanata by taking the advantages of version. Create a new version in Zanata, named as ibm-translation. Ask IBM to upload po files there. Translators could review in the web UI, or review offline by downloading the po files. Translators make comments or updates through Zanata. IBM make improvement based on the comments. When translators are OK with IBM translations, merge these two versions. Option #4: No pre-review at all Some translation teams don't mind it much as long as it covers only the missing areas of existing translation. I think, #3 is the best one. I have made a test in Zanata test server. I created a new version of Nova, see: https://translate-dev.openstack.org/iteration/view/nova/ibm-translation I asked IBM to upload the Italian translations of document "nova". You could see the percentage increase to 90%, while in master version the percentage is still 37%. As translation team becomes more and more popular, there would be more people (or company) who want to contribute their translations in po files. I want to create a formal process for those who want to contribute translations in batch. So I would like to collect your feedback to option #3. @ Team coordinators, let me know what do you want to check in pre-review, and whether using Zanata version could support you well. @ Clark and Elizabeth, let me know if you are OK with it from the infrastructure perspective. @ Carlos, maybe you could give us good suggestion as a Zanata expert. Thank you all. Best regards Ying Chun Guo (Daisy) Douglas Fish <drfish@us.ibm.com> wrote on 09/10/2015 08:01:11 PM: period that that
Hello Daisy, all, Thanks for translation contribution by IBM.
Very welcome!
Agree.
As far as Japanese, please directly email them to me as attachment. I will pre-review and upload them. It doesn't take so long time. Just a quick pre-review. Then, openly review by Japanese team in Zanata.
I feel #3 is a little complex. As a result, #3 is a burden to the reviewers. So, I think "#4': almost not pre-review" is reasonable. Thanks, KATO Tomoyuki
ceilometer, glance, heat, nova, ironic, neutron, cinder, keystone, and swift.
BTW, openstack client is important :)
On Fri, Sep 11, 2015, at 11:00 AM, Ying Chun Guo wrote: provide tooling to perform the translations upstream. This is where translations should happen instead of happening at many different downstreams who then have to figure out how to merge the translations when they are pushed in batches. We should just be translating as we go and working together. If there are deficiencies in the tooling that prevent this from happening we should work to fix them not work around around them by suggesting different downstreams should solve these problems individually on their own. All that said, you can edit translations locally in po files and push them up with the zanata cli tool right? If translators prefer that workflow to the web UI they can do that today unless I misunderstand something.
Hi Clark, On 15/09/15 03:04, Clark Boylan wrote:
Just quick context here. Quite a lot of translation of software around the place is not done "in house". Rather, a file gets sent over to a contracted professional translation agency. These organisations have their own workflows and based on my experience would politely decline a request to use different tooling for a single client. Where the translation is being performed by organisations well integrated with the community, for sure - let's get it done 'as we go' and make it well visible for all the good reasons you've listed. However, I do think we want to ensure that if someone's going to foot the bill for a serious advance in a language's translation completion, we don't scare them away by making unreasonable demands of their contractor :) Regards, Tom
Hi, Generally speaking, I welcome *THIS* IBM contributions, but I have some concerns on doing this at this stage. Each translation team has their own translation guideline and it may be different from IBM guideline. The guideline should be discussed openly. The second point is the limitation of time to review. Most translators and coordinators are busy translating strings for Liberty, and they don't have enough time to review more. It depends on a situation of each language team. A final decision should be delegated to each language team. Some language team may accept IBM contributions for Liberty, but some team may defer them to Mitaka. On the other hand, I am not a fan to making PO file contribution a general process. I don't would like to encourage to push whole PO files to the community. Why don't you make translation in the upstream? There are several reasons. - There are other translators in our community and PO file translation leads to translation conflict. We should avoid duplicated work. - Translation guideines. I believe existing language teams have some guidelines for terminologies, which may be different from a company guideline. Pushing whole PO files may force translators to review bursty and if there is a mismatch in guidelines not a small number of efforts need to be changed. - Reviews and discussions should happen openly in a language team. A translation guideline in a company may be different from one of the community. In such case, internal reviews contribute less. Thanks, Akihiro 2015-09-12 3:00 GMT+09:00 Ying Chun Guo <guoyingc@cn.ibm.com>:
participants (6)
-
Akihiro Motoki
-
Clark Boylan
-
Douglas Fish
-
KATO Tomoyuki
-
Tom Fifield
-
Ying Chun Guo