Question

I’m doing some i18n on a web-based app using Django, which uses gettext as its i18n foundation. It seems like an obvious idea that translations should be stored in the database, and not difficult to do, but po files on the filesystem are still being used. Why is this?

My current suspicion is that the benefits of developing a db backaged are simply outweighed by the reliability/familiarity of gettext as a well-established package. Are there other significant reasons for continuing to store the translations on the filesystem?

Answer 1

Performance is the main reason. Gettext is not using a database because a database will always be considerably slower than a file. The load time of the dictionary is very important and for this reason almost everyone is using files for that.

Also, the compiled gettext files (.mo) are optimized for loading in memory and for this reason they are more appropriate than plain text files (like not-compiled .po files).

You can always use translation platform, probably that uses a database backend, for doing the translation and export the results to text files. Examples: Pootle, Narro, Launchpad Rosetta, Transifex (hosted only).

Do not confuse your application language files with the localization database. Your application should use file based dictionaries that are fast to load and your localization system probably will have to use a database and logically be able to export data to files.

By the way, using gettext is probably the best technological decision you may be able to make regarding localization. I never seen any commercial solution or in-house developed to be able to compete with it on features, tools and even support.