Internationalising time estimate



  • Here's the problem: I was given the responsibility of getting our 100-or so page website internationalised. Maybe a quarter of the visible text on the site was gettexted (ie, was wrapped in _()). I had no experience whatsoever with gettext or any else to do with internationalisation. The experience of the 2 other developers was almost 0. I was the only person working on this, and it was only a part of my job. Now, 3 months later, we have:


    • Tool written from scratch for online translation - basically directly modifies the translation (.po) file (probably not ideal, but that's a different wtf)
    • 2 'complete' languages (in production, with the occasional English showing, mostly were text has changed)
    • 3 almost complete languages (within about 100 strings of being translated, or various states of proofreading)
    • 2 other languages being translated
    • Search in each language
    • Dynamically translated details (keywords etc)



      There are still a few issues unresolved, but currently only maybe 5 places where there's English (that I know about) - and none are on the main pages.



      I'm now facing losing my job because this has "dragged on for so long". Assuming 1 person working somewhere between half- and quarter-time on it, is 3 months a reasonable amount of time to have taken? Should / could I have been at this point in a month (including discovering all the wonderful quirks of gettext, and writing the tool)? There was no indication at the beginning of a deadline or an estimate - a mistake on my part. Any comments are welcome.


  • inb4 "Screw 'em!   You're better off not working for them!"



  • Fsck them. They should have given you a reasonable schedule in advance. Internationalising is an issue that can take any amount of time, depending on how careful you do it.



  • You probebly shouldn't have written your own tool, unless that was part of the plan. One of the biggest plusses about gettext is that it is a established standard. Meaning for one, that there are already tools out there to work with po files, for all platforms, probebly also web apps.

    Then the actual translation is a time vs. perfection problem. Which takes longer the closer you get to 100% translated. A strategy that saves time and thus money is to simply do a best effort 99% translation, and add a button on the website to report non-translated items. I also think there might be tools out there that can scan code for loose strings, but i am not sure and those also don't do 100% of course.

    A often forgotten problem about translation however is that different languages need different ammounts of space to convey a message. So depending on the layout, you could get some issues in that area which can suck up your time. German for instance is a language to be feared, from memory it often needs 15% more text to convey the same message. I am sure the stats are out there, so correct me if i am wrong. 

    Another problem would be legacy applications that where written to work with latin-1 or some other non-UTF8 encoding. Checking and converting everything to correctly handle UTF8 can be a bitch if you have to do it after a project has finished.

     

     -edit-

    I also agree with those above. Fuck them.  Either you are a developer or you are a manager. You simply can not be both wrapped into a superman package. If you are a developer working on this project without a manager, then they shouldn't be surprised it wasn't (time) managed. 



  • And don't forget number format, time format, date format. And all those pesky error messages that might pop up when something goes wrong. For non-western languages, unicode and right-to-left might make it even more funny.



  •  "That's right," the man said. "I couldn't remember the word." He was the only t, then high school students, and, finally, to anyone aged 13 and over. The website currently has more than 175 million active users  in amount of visitors, making Facebook the most popular social network, followed by MySpace and Twitter.other human at the loading dock this morning. The man didn't have a name, just a number, like the rest of the robots. Paris, at Night.


Log in to reply