I'm seriously considering a change in how Wget translations are distributed: I want to uncouple translations from the source. First I'll explain what I have in mind, and then I'll give my reasons. I would like to hear whether you think this is a good idea, and why.
What: Uncouple the translations from the source in such a way that the *.po and the related *.gmo files are no longer distributed in the `wget-VERSION.tar.gz' archive. The po/ subdirectory of `wget-VERSION.tar.gz' will still contain the Makefiles and the infrastructure for NLS, including the `wget.pot' file -- in other words, everything except the actual translations. The translations will be shipped in a separate archive named `wget-VERSION-translations-N.tar.gz', in the same directory with the source archive. VERSION is the release of Wget, and N is the version of the translations distribution that corresponds to that Wget release, beginning with one and incremented with each new version of the translations tarball. The translations will only contain the `po/' directory and will unpack to `wget-VERSION/po/'. It will also contain MO files compiled with GNU gettext. In other words, unpacking both tarballs in the same location will produce the result equivalent to what you get after unpacking the current `wget-VERSION.tar.gz'. Why: * Relieve the maintainer of the burden of handling the translations. It will be possible for the translations to be handled by a volunteer, or by the Translation Project, if it so chooses. Or a new translation tarball could even be created by having a script wget the available files from the TP and create the distribution automatically. * Give the translators more time to translate the packages, while allowing the faster ones to react immediately. For example, the first version of the translations tarball can be issued one week after the release, the second version a week after that, and then as a new translation arrives. This means that the first version of the translation tarball will already contain the input from the faster translators. But unlike the traditional model, this will not leave users of languages with slower translators empty-handed; they simply need to wait for a week or two until their translation appears in the tarball. * Remove the need for a "string freeze", which can cause problems when very important messages need to be changed, and *still* don't guarantee that all the translators will have enough time to translate the package. * Allow the users a simple way to update the translations on a running system. * Allow the users who don't need or want I18N (sometimes referred to as "Americans") to download Wget without getting the translations at all. If they change their minds, they can still get the translations later, and install them. Q: I don't agree with this. Is your decision final, or will you take counter arguments into account? My decision is not final, and I would definitely like to encourage a discussion. Getting feedback about this from Wget users is one reason for sending this message. Q: Why change what has always worked? Because I am dissatisfied with how it has worked. Integrating PO files into the source tree detracts time that could be spent on development or answering bug reports. Some people find this loss negligible or acceptable, but I think differently. Besides, see the points above; each of them addresses an issue raised by users and developers over time. Q: But everyone else does it differently... Most of them do. But then again, the best way to effect a change is to lead the way. Q: But won't this make the translations harder to come by, and hence hurt the Translation Project? I don't think it will make a shred of difference. There are people who care about translations. There are people who don't. The latter will not download the translations, and will be none the worse for it. The former will either download the translation, or notice them missing and *then* download them from the same place they downloaded Wget from.
