How would that work in practice? By its very nature open data would also be accessible to proprietary programs although the reverse need not be the case.
Seems like we will need open data and code. Either one open will not do.
That is a good thing. If program X works best on my data I want to use it. There are a few examples where people do mix programs from different companies. Musicians use MIDI to connect their favorite keyboard to a synthesizer from a different company all the time - sure it is tied to hardware, but it need not be and is a perfect example of what should be possible for any user data: mix and match.
> although the reverse need not be the case.
It doesn't have to be, but if users demand it, it will be.
> Seems like we will need open data and code. Either one open will not do.
Open data means we can create the code. Closed data is a lot harder to deal with than closed code.
When your data is complex, the processing done to it will be complex, especially if you need to guarantee invariants (eg. referential integrity or database constraints).
Edit: I envision it a bit like streams of data that you can subscribe / push to. Think RSS mixed with a pub/sub type of model. You would subscribe to the hackernews datastream, and submitting articles and comments are done using push. The push message would have some predefined metadata fields that are obligatory (article url, title, summary or comment text).
Can you imagine if Android licensed iMessage instead of building Hangouts? Yes, we’d all be texting on the same protocol, and yes we’d have a choice of clients, but at what cost?