I’ve been looking at potentially replacing the current string array based translation code with something more mainstream. One obvious candidate is gettext.
gettext has been around for a long time and is used in many applications. One particular downside is that it generally works with a single file for a given language/locale. Now this is great for an application where you have control over all strings. Obviously, this is not the case for ZenMagick. Looking a bit around showed that there are various solutions to handle this.
One particular good solutions is pomo. pomo is a thin layer on top of gettext. It allows to have multiple translation sources (which can be merged). Furthermore, and that might be the main benefit, it is pure PHP, so even works on systems where gettext is not installed.

One prominent application using pomo is Wordpress. I’ll have a play with it and try to write about any progress.