Ok, before this gets closed, I'll give you my thoughts on the matter. I believe that ISPs already implement everything in the IETFs zero-conf specification, using traditional DHCP and DNS. In the worst case (a completely unmanaged network) AVAHI is often sufficient.
The thing you seem to want, is service discovery, which is quite possible, but what you are missing is the matter of control. The service operator should always be in charge of deciding what to make discoverable, and how to do it. It is not the ISPs right to expose the services on my network at a pure TCP/IP level.
Secondly, the Internet at large is too large to ever be "discoverable" in a meaningful way, unless resolution is scoped, probably geographically. On the local level, the only reason that Bonjour works is that its scoped to a LAN. If iTunes ever even considered sharing my music with my neighbors, I would be forced to slay it mightily. Bonjour was made for people who can't be bothered to configure either end of a service provisioning.
Using the WWW as an overlay, server admins and developers can implement services over the web that are discoverable, using a wide array of underlying technologies to implement standards for discoverable services. We can discover them with search engines, and access them with whatever applications we write or otherwise obtain which were designed to consume their provisions. Fedex is free to manage the many printers in many differant locations, all of which are in some state (idle, printing, queueing, etc) at any given instant using software that is designed and configured as They want it, not the ISP.
If Fedex wanted to set up web printing, they could do so now, and you could print to them using the MS Internet Printing drivers to do so. line printing comms protocols were never really designed to provide the robustness you need to hit a printer on the other side of the world anyway. New drivers and protocols would be needed to directly access the device in the same way they do one connected via USB. Your OS might choose to show it to you, and make it look identical to a local printer, or they might not.
So, in sum, I would argue that we already have all the capability to do what you want if we choose to do it, and that freedom to do so or not is essential to keeping the Internet free and democratic. Placing the responsibility at the ISP level is inappropriate, and would cause wide divergence between ISPs in ways that really should be predictable commodities. imagine if your service was only available on one ISP, or even 4 out of every 5. It would be pandemonium, at first, until the entire internet degraded into sets of stagnating walled gardens like Apples. The internet cannot be allowed to become a cable package.
“How practical would it be for, say, an ISP's provider to develop a Bonjour-like protocol for the Internet?” If you have ever worked with office LANs you know how much of a headache the networking traffic caused by Bonjour-like protocols are. It literally saturates networks to the point of making it unusable. Fine in a small home network though since you will never really reach the scale where it becomes an issue at home. But many places actually disable the broadcast traffic on their networks from the switch/router to kill this stuff. On an Internet scale? 100% no way. – JakeGould – 2015-03-13T05:17:51.360
@JakeGould: Bonjour doesn't use broadcast, and AFAIK it was specifically designed to let clients cache everything. Besides, just how many DNS queries exactly you need to saturate 100 Mbps? (If it was NetBIOS on the other hand, yes, I could believe it being painful.) – user1686 – 2015-03-13T09:21:05.347