Tuesday, February 21, 2012

What will be the global benefit of using #Git for #MediaWiki

Git is a version control system. The Wikimedia Foundation is going to use it and frankly, I have no clue why. What I do understand is that the waving around of hands and the exclamation of buzz words looks good. I am not aware of the existence of a cost benefit analysis, something that would make things understandable.

My impression is that because of the introduction of git, the quality of the localisations and the internationalisations of MediaWiki and its extensions will be negatively impacted.

As it is, changes committed to Subversion, the current version control system, result in new messages in translatewiki.net within a day. These new messages become available to the twn community and as a result there is a steady stream of messages that need textual improvement and/or an explanation. As development progresses, the messages improve and the localisation is under way. At the time new functionality is ready for deployment it is localised in many languages.

With Git, there is no such thing as new messages that make their way to translatewiki.net. As I understand it this will only happen when the developer is done, the software is reviewed and is moved into "production" status. This means no internationalisation review, no textual improvements, no help texts and no localisations once software goes into production.

When a translator costs one hundredth part of a developer and we support three hundred languages then reducing the cost of development is not necessarily where we make our big wins. Our language support benefits from the current model and I fail to see why we should git going.

Really, what is the global benefit from using git over Subversion ?


Sumana Harihareswara said...

People who wish to know why MediaWiki development is switching to Git may wish to read:


tbleher said...

As I understand it, the biggest bottleneck in developing MediaWiki right now is the numbers of reviewers. Git promises to ease this problem, by switching to a pre-commit review model, where reviewers can send back problematic commits to be reworked, before they hit trunk and have to be reverted. Sp I expect the development of MediaWiki to speed up once Git is properly set up.

I see no fundamental reason why translations have to be delayed the way you describe it. I think it will probably take some time to set up everything correctly and to adapt to the "new way", so please be patient :)

(This is my personal view as someone who has used Git a lot and follows MediaWiki development via wikitech-l. I'm not actively involved in the Git migration, so take this with a grain of salt.)

GerardM said...

Setting up the internationalisation and localisation process only after migrating to git means that the fact that we support 300 languages live is considered to be of secondary relevance.

Once the conversion is done to git there will be a considerable time of upheaval and consequently resources will be slim.