Will the cost of maintaining the hand-rolled code exceed the cost of working with XMPP tech (and subsequently benefiting from an existing community and proven reliable code)? Hiring someone for an existing standard, for example, means less legwork documenting and codifying your own system.
What rate of feature growth do you see, if any? Whatever solution you choose now, assumptions will start being made against it in future code, and they will slow down a switch later.
One way you could research this further is to do the bare minimum test case of both techniques in isolation from your other work. You won't have the "but I have to consider x" in the back of your head, doing it that way.
Will the cost of maintaining the hand-rolled code exceed the cost of working with XMPP tech (and subsequently benefiting from an existing community and proven reliable code)? Hiring someone for an existing standard, for example, means less legwork documenting and codifying your own system.
What rate of feature growth do you see, if any? Whatever solution you choose now, assumptions will start being made against it in future code, and they will slow down a switch later.
One way you could research this further is to do the bare minimum test case of both techniques in isolation from your other work. You won't have the "but I have to consider x" in the back of your head, doing it that way.