One important point to understand about software translations, in contrast to translations of other text materials, is that translators will often have a rather hard time getting a complete picture of what they are translating. Software translations are necessarily based on rather short fragments of text: Every label you give to an <option> in a <radio>, every string that you mark for translation in an i18n()-function call, will form a separate "translation unit". In essence, each such fragment will be presented to the translator in isolation. Well, not complete isolation, as we do try to provide translator with as much meaningful context as can be extracted, automatically. But at some points translators will need additional context to make sense of a string, especially where strings are short.