This is a bug in jQuery at least, so maybe technically I'd want to get the fix upstream? I don't know, but I found the effects here.
While the node.js engine seems fine with taking objects and arrays when making requests without serializing them (or at least serializing them properly), it seems that jQuery (2.1.4?) is not so forgiving--it seems to be some kind of serializing bug where the "[" and "]" for the list appear in the wrong spot, on the field side instead of the value side, destroying the request as intended (this bad input can be seen as it is fine when it leaves the manager, transforms inside jQuery (I haven't caught exactly where yet), and bad when it gets to barista). In our case, since groups (provided-by) is an array at the top level, we would need to encode it to make sure is survives the trip to the browser (as above); however, for the time being, Minerva cannot decode that string (expecting json object as-is), so we are blocked. Needless to say, this was causing some really weird bugs. Would a jQuery upgrade help here?
Since I'm serializing the requests as a string, and minerva seems happy to decode them as such, they get through fine.
As a temporary workaround, we are just going to take the first provided-by argument as pass it through.
This is a bug in jQuery at least, so maybe technically I'd want to get the fix upstream? I don't know, but I found the effects here.
While the node.js engine seems fine with taking objects and arrays when making requests without serializing them (or at least serializing them properly), it seems that jQuery (2.1.4?) is not so forgiving--it seems to be some kind of serializing bug where the "[" and "]" for the list appear in the wrong spot, on the field side instead of the value side, destroying the request as intended (this bad input can be seen as it is fine when it leaves the manager, transforms inside jQuery (I haven't caught exactly where yet), and bad when it gets to barista). In our case, since groups (provided-by) is an array at the top level, we would need to encode it to make sure is survives the trip to the browser (as above); however, for the time being, Minerva cannot decode that string (expecting json object as-is), so we are blocked. Needless to say, this was causing some really weird bugs. Would a jQuery upgrade help here?
Since I'm serializing the requests as a string, and minerva seems happy to decode them as such, they get through fine.
As a temporary workaround, we are just going to take the first provided-by argument as pass it through.