Discussion: how to distinguish between +end (lookup) and +end (distribute)?

As I’m implementing a simple chat client I’m finding some holes in the clj-telehash implementation.

A blocking problem at this point is how to distinguish between +end (lookup) and +end (distribute). This problem occurs when a switch sends an +end that is not its own IP:PORT but an application-specific end (e.g. user@chat).

The fundamental problem is that the receiving switch cannot know is the +end is for lookup purposes or for distribution purposes. The +end would be for lookup purposes if the counterparty application wants to find one of its chat contacts (e.g. jer@chat). The +end would be for distribution purposes if the counterparty application wants to distribute its own chat id (e.g. steven@chat).

If the +end is for distribution purposes it should be added to the kbuckets since when another +end arrives for lookup purposes the counterparty’s IP:PORT must be returned in .see . If the +end is for lookup purposes it should not be added to the kbuckets since we don’t want to send out an IP:PORT with .see for a switch that’s guaranteed not to know about the +end it is looking for. Moreover, there is only one slot per end in the kbuckets. If we add each end to the kbuckets we’ll have multiple switches per end. Is this as intended?

Am I missing something here?