Your data on German is not entirely correct - as in my example of the "translate" verb in the i18n discussion, there are nouns derived from Adjectives (das Englische, die Angestellte, etc.) that have case-based adjectival inflection ("ins Englische" — into english / "aus dem Englischen" — from english). While in this case you can use prepositions for role assignment, someone still has to deal with the noun modification.

There are other examples of noun modification in German, the genitive case ("der Vater" => "des Vaters") which is somewhat equivalent to english posessive, and a completely irregular plural — but fortunately it doesn't look like those are relevant for Ubiquity.

The other thing is, that in strongly case-based languages — like your polish example — somebody has to deal with the noun modification resulting from the case, even in those cases where there is an alternative way to identify role of arguments. Which leads me to the third thing:

Russian does not actually have an natural alternative to case-marking in all cases. Consider "search for hello using google" (I am going to use a russian "romanji" equivalent because my keyboard is not configured for cyrillic):

naidi hello guglem

or

naidi guglem hello

alternative:

naidi hello s pomosh'y guglya

literal translation "find hello with the help of google", and about as natural for a russian to think of and type as that. The same applies for all "with some tool" type of expressions.

To be fair, if one can think of google as a place where you search for something, as opposed to a means/tool for the search, one might be able to use "naidi hello v gugle" instead (as in "find hello in google" or "search google for hello"), but that might not be possible with each and every tool out there. In some cases, case is really the only way to go.

However specifically in case of limited number of nouns, like a list of search engines, the list itself could easily serve as a role identifier, and there would not be a need to use an adposition.

Anyway, I think there’s a mistake on the example c).

‘hon-i’ is incorrect, as ‘hon’ doesn’t exist as a demonstrative word (even as a word) so it doesn’t mean ‘this’. I think it’s been written like this just to mark the separation between the “word” and the suffix.

The translation for ‘this’ is ‘hau’, so the example may show ‘hau-ri’, which ends up in ‘honi’.

I think there's a little mistake in your German example though: den großen Hund the=NOM big dog

"den" is not nominative, it's accusative case.

When talking about this stuff with Jono, I have always been caught up on point one, about the case of a particular item being difficult to identify. I was especially hung up on those situations where declining the noun was difficult for computers (irregular markings, ablaut, etc.). As for the point about delimiting the whole phrasal argument, I had never thought of that before, and it looks quite problematic.

For the third point, I would suggest picking the declension that would be most likely to occur in a Ubiquity command in general. I would guess this would be accusative for nom/acc languages, and absolutive for ergative/absolutive languages. For example, let's say your language has an instrumental case, and you ask Ubiquity to translate something using Google. Ubiquity would represent Google.instrumental as two words: an abstract instrumental "preposition" and Google inflected for the accusative.