The Jed Internationalization Toolkit
The Dojo Foundation and Future
I’m excited that The Jed Toolkit has been accepted into the Dojo Foundation so its users can be sure that it will be a safe, unencumbered resource for them into the future. I’m extremely happy to be part of a family that includes require.js, sizzle, and the dojo toolkit.
I was drawn to this problem because there were so few people considering its intricacies, but I was shown by some very smart folks that there was a lot more to internationalization than the little library I wrote. So I wrote another library, and ported a few others.
I was quite happy with how these were turning out. They weren’t especially hard to create, because most of them follow well-documented specifications. I really liked how ICU MessageFormat made a lot of decisions based on how translators think, instead of how programmers think. But naturally, they locked away that goodness behind a syntax/grammar that no non-programmer should ever have to deal with. MessageFormat is great for translations but not for translators. Not in the real world at least. That’s when I realized that the problem was not (only) the tools for writing international apps, but even deeper: in the tools and integration with translators.
It’s all about tools
The translation space hasn’t grown much since computers first existed. We can barely encode files correctly in 2012. However, in other spaces, like content authoring, we have a whole system of tools and integrations to map non-technical users’ intent to structured usable data for consumption. If your local tech writer wants to start a blog, they can! They don’t need to know how to set up a server or that HTML even exists.
The process of getting an app translated is cumbersome, and is a blocker to getting good applications out there. FTP zips and crazy XML specs mixed with Word Documents rule the landscape. There are no decent apis, or automatic integrations that anybody is using at scale. I want to set out to change this.
Translators aren’t all to blame. If you were a translator and got the message “fair”, would you translate it as a carnival, or as ‘just’? We set our translators up for failure with our context. We can do better. We can describe messages, and their variables. We can offer examples and photos of the context. We can even translate the app in real time and they can see their translation literally running in the place it will live.
The goal of the Jed I18n Toolkit is to help make the internationalization process much more accurate and enjoyable for all parties. We should be able to write our messages directly in our templates in whatever format we think is best. Our messages should be automatically culled, and deduped and sent into a translation queue. The translator shouldn’t be presented with anything other than things that help them translate. The programmer’s format should be irrelevant. Context is king, and a bunch of crazy sprintf characters and html are just noise. When the translations are done, they should exist as a service or api and be updated in real time. Gone should be the days of the 2 month translation code freeze. You should be able to write a post commit hook that gets your translations through the system as fast as you can find someone to translate them.
There’s a lot to decide on how to bring all of these ideas into the project in a generic, but still usable way, and it will take some time to get everything right. Right now I’m starting by putting in the few open source projects that are already out there as well as showing early beta work on some of the integration tools. Please be patient with me and send me your suggestions and frustrations so we can finally bring internationalization out of the dark ages.