Request management: cache/abort promises

[juanca]

There are some edge cases were promises need to be managed: invoking two fetches from the same collection requires aborting the first one (i.e. only have the last fetch in-flight).

It would be nice to have a collection / model instance cache promises and return them when appropriate. Or, abort previous promises.

I'm not sure which is the right solution at the moment. Will sit on it. Would appreciate thoughts / opinions.

[juanca]

I think each time fetch is invoked, a unique fetch key should also be provided. This key is used to store / lookup a fetch value (read: XHR), cancel it and clobber it with a new one.

Unfortunately, it doesn't make sense to do it when brainstem-js is so tightly coupled with backbone (since it'll just return the XHR and this logic can be done on a per-component basis).

It makes sense to implement this when brainstem-js is de-coupled (server parsing / request formatting, isomorphic fetch(?)) and everything backbone-related in brainstem-backbone. This way brainstem-redux can just consume brainstem-js (and have all those nifty things?)