I don't understand - if Zotero doesn't have the data how could it create the reference?

sorry, you need to be more explicit about the workflow you're imagining - right now I don't see exactly what you mean - but Zotero can't just "get" data from any webpage. The page needs to present the data in a systematic fashion and someone needs to write a translator.

This just isn't going to be possible. Zotero works with items in its database. It has a robust style engine that lets it create bibliographies in hundreds of different ways, but its core is really keeping track of these items. If you just need a way to turn bibliographic data into citations, you may want to look into using citeproc outside of Zotero-- citeproc-js, the implementation of CSL 1.0, can be run using Rhino or another JavaScript interpreter. The earlier version of CSL has implementations in Python and Haskell as well.

If you most of all want to employ the OpenOffice plugin, you will definitely need to make Zotero items for each entry first. You could simplify that process by writing a site or import translator, but the items need to be in Zotero before you cite them.

Thanks ajlyon, but why can't it create a temporary database item which it drops when it has given me the bibliography entry? (My computer will not mind about the extra work.)

I guess that's possible in theory - I think the usage case is so rare, though, that you won't see anyone actually coding that - so if you really feel the time saved is worth it you'll have to write a script yourself or find someone who does that for you.

I find it hard to believe that there is actually a situation where it wouldn't be preferable to save a couple of items in Zotero and then drag the whole batch over to whatever alternative you're using, but that's up to you.