A service hub is a good definitions of this project. The xmpp protocol handle the SSL communications and prevent address spoofing already, of course each server would have their certificate, we are not trying to reinvent the wheel.

I think a lot about security of the whole system and for the BIP32 part although it's an elegant solution it must not belong to the server to take it because I think the strongest for security is that each client can sign every bit of information they transmit with a unique, publicly identified Bitcoin address that any willing client can verify to correspond with the intended entity. And for this the client software must generate a list of 100 addresses in advance, sign each of it and send it to the server for storing it and repeating this when addresses are used. Of course we can sign only a public masterkey and clients can verify it... but then anyone can tell all the future addresses resulting in a loss of privacy.

A good compromise is that regularly the client publish and sign a new public masterkey this way we have a strong scheme where even a corrupted server cannot forge addresses because of the signature and by repeatedly generating new masterkey we maintain a good level of privacy. I have already framed the majority of this protocol and a very early pre-alpha (proof of concept) implementation in python will be published on github very soon...

Yeah Bitcoin-Qt is very hungry on resources, electrum is what I intend for the majority of users and running his own electrum server on a VPS would be great and can already be made. For this thing I will publish the client and the server part so you can have your own server on your own domain if you want... It's a pretty light protocol on resources and every merchant sites can have their own server installed, so maximum security for customers and a nice business payment address with their own domain.