From: Richard Conroy <richard.conroy@...>
Date: 2007-10-09T20:07:42+09:00
Subject: Re: [OT] Translation Service

On 10/9/07, Clifford Heath <no@spam.please.net> wrote:
> Richard Conroy wrote:
> > I am testing the internationalization strategy, not the end release to
> > users in whatever language.
> > [...]
> > All of these are testable with machine translations.
>
> Translation is not required for testing - you can just use lorem ipsum,
> generated into bidir or other characters to the length you might need.
> We tend to bracket each snippet with characters that are visually
> recognisable, like [ ], to aid in manual verification that the
> translated text is being used and not the default English text.

True. I would actually end up doing something like that because it is
then something that I can link into automated tests - scraping the
web UI for a regular expression is wonderfully simple with tools like
WATIR.

However, I know people in our company would be more confident with
something close to the bare metal (simplified chinese or equivalent).
Its close to our existing testing methods. It means that they don't have to
take a developers word for it (me). It also means that people other than
me can make a decision to pay for human translations for their own purposes
(get reviews/buy in from regional domain experts etc.).

So yeah, both needs are important, and simple watermarking of english
resources and using made-up locales is actually a better form of testing.
But actually making real translations means that nobody questions it.
Real translations (machine) are handy for other purposes too.