I remember a time when, as a junior developer, I'd log into operations, and hand the operator my tape reel with my work for the day - and she'd go off, mount the spool, and start the streaming, until my files were available in local storage for a few hours, and then .. when logging out, I'd have the operator re-stream the files, and off we go ..

So, why not use space as a large tape buffer? Send the satellites off, 20 minutes of light-speed away, and use the big gap as a cache.

Seems sorta feasible to me, but I guess I'm overlooking some ridiculously obvious factor, such as launch costs, viability, etc. But really, seems like a fleet of satellites spread out around the solar system could function as quite a nice medium ..

Using your example, and assuming you want to send a message to a moon base on the surface of Callisto, a moon of Jupiter: while you may have a clear line to Jupiter, the point on the surface of Callisto that you want to send a message may not be in your line-of-sight because it is facing away from you. So your route becomes Earth -> Jupiter Communications Satellite 3 -> Callisto ComSat 10 -> Callisto Base Alpha

Assuming you want to send a message to the far side of Luna, even then the best path may be a commsat in high orbit around Luna -> Comsat in low orbit around Luna -> FarSide Base.

I agree, rarely are you going to need to bounce from Mars to Jupiter, but for low-priority bulk traffic (scientific images, etc) and enough nodes, it might be worthwhile to bounce from Mars to Jupiter to Luna to Earth if all communication nodes were otherwise idle. The Delay Tolerant Network Vint Cerf is working on supports knowledge of when certain antennas have available communication windows to different nodes, and can schedule and route communication accordingly.

Source: Myself, did a presentation on it in college

Let's call the amount of transmit energy needed to get a signal from Earth to Mars with it having a specified strength at the receiver 1 unit of energy.

Earth to Jupiter would require about 64 units of energy, so from the sender's point of view relaying through Mars saves them a lot of transmit energy.

The transmitter on Mars would need about 49 units of energy to get the signal to Jupiter, bringing the total to about 50 units to get from Earth to Mars.

Whether or not spending 1 unit on Earth and 49 on Mars to relay is actually cheaper than spending 64 units on Earth to go directly will depend on the cost of energy on Mars and Earth.

Furthermore, even if you have clear line of sight, repeaters on the way can be helpful because the inverse square law even holds for lasers (the light isn't perfectly parallel, there is no such solution to the maxwell equations (outside of an infinite plane emitting light perpendicular to it), instead it's a gaussian beam).

Last, you only have a limited amount of spectrum you can send to or receive from a specific point in the sky. Beyond that, it gets hard really quick. If you have repeater stations far away so that they can be targeted from the distance, you can increase the maximum bandwidth. This is mainly a concern once the interplanetary bandwidth is maxed out.

If your router decides it has a direct route to Jupiter then it probably won't hop to Mars.

On the other hand, the latency penalty might be quite steep. There is probably a latency/bandwidth tradeoff that needs to be solved for every packet. I think today's internet has the same set of tradeoffs (plus some, like cost). If packets could carry their own metrics, routers could make better-informed choices.