There's an interesting history of one-way content delivery by satellite. Although info on the service is now sparse, since at least 1995 PageSat was delivering a constant feed of Usenet updates via satellite broadcast as a means for smaller ISPs and BBS operators to avoid the (at the time) substantial bandwidth required to receive NNTP - I once worked in a rural university IT operation that no longer used the service but still had the hardware lying around. This sparked an interest in the broader ways that content has been delivered by satellite, predating the practicality of two-way VSATs. A prime example is the use of IP-over-MPEG to deliver the downlink side of a consumer internet connection via satellite television broadcast, while the uplink uses dial-up or ADSL. This arrangement is unusual in the US today but still in use occasionally.
On the other hand, there are still organizations trying to use the same "broadcast internet" model. Cidera nèe SkyCache flopped on using satellite broadcast to deliver content to edge CDN nodes in the early 2000s, and Outernet is currently flopping (minor opinion inserted there) on delivering educational content using a similar model.
This is viable in certain niches. The broadcast industry, for example, makes use of similar systems with the Public Radio Satellite System (affiliated with NPR) using one-way UDP broadcast to deliver live events and recorded syndicated programs to member stations. This has the upside of lower and latency and much lower jitter compared to the internet but the margin of superiority over conventional internet delivery is getting narrower by the day.
On a slightly related tangent, the general nature of the design of satellite data relay (with spot antennas covering relatively broad areas of the planet) means that the downlink channel of any satellite internet customer can generally be received over a wide area, even if they are using a modern two-way VSAT. With reasonable encryption this doesn't pose much of a problem, but some satellite ISPs, particularly in the developing world, continue to make use of unencrypted IP-over-MPEG for their downlink. This allows a malicious eavesdropper anywhere in the region to receive the traffic one direction, which has the extremely important implication that it allows them to hijack TCP sessions by intercepting sequence numbers - this enables IP spoofing over TCP, something that is generally regarded as impossible. This is sufficiently reliable that it has been used as part of the C2 infrastructure in botnets. I have a research paper on this somewhere.
satellite is still the only way to get internet in many regions of Australia. This is a major consideration when it comes to how websites are designed if you intend for that audience to be able to access your service. Tiny uplink bandwidth, large amount of latency.
Therefore, for instance, large images aren't as big a problem as many many small images, or transitive dependencies.
> Tiny uplink bandwidth, large amount of latency. Therefore, for instance, large images aren't as big a problem as many many small images, or transitive dependencies.
Is this a typo, or is it just that many small/transitive dependencies are much worse?
many small transitive dependencies are much worse when request turnaround is up to 1 second. so for an external image loaded from an external CSS file, that's 2 seconds. It's a bandwidth scenario where reducing the number of seperate http requests makes a huge difference.
An unreleased SNES game, Cooly Skunk, was recently discovered and likely never would have come to light if it wasn't for Satellaview broadcasting a demo of the game while the service was active, and a user preserving it on a Satellaview download cartridge. Fascinating stuff.
A good deal of the Satellaview games we do have preserved in this manner are incomplete. The BS Fire Emblem games and the LoZ remake, for instance, both lack sound. Supposedly the BS Zelda games had "live" narration that was also being streamed over; that could not be captured, so we'll never know what that sounded like (unless someone happened to be recording their playthrough at the time on a VCR or something).
I really love that Nintendo did something like Satellaview, and it's always been something that has struck me as a niche product that was far ahead of its time. It's too bad that we will never be able to experience what using the Satellaview was like (admittedly, probably pretty slow and crappy), but at the time I can only imagine it would have felt quite novel.
Living the 80s-90s in Japan was like living 10 years in the future, it’s crazy how much technology was already making its way into the consumer market.
On the other hand, there are still organizations trying to use the same "broadcast internet" model. Cidera nèe SkyCache flopped on using satellite broadcast to deliver content to edge CDN nodes in the early 2000s, and Outernet is currently flopping (minor opinion inserted there) on delivering educational content using a similar model.
This is viable in certain niches. The broadcast industry, for example, makes use of similar systems with the Public Radio Satellite System (affiliated with NPR) using one-way UDP broadcast to deliver live events and recorded syndicated programs to member stations. This has the upside of lower and latency and much lower jitter compared to the internet but the margin of superiority over conventional internet delivery is getting narrower by the day.
On a slightly related tangent, the general nature of the design of satellite data relay (with spot antennas covering relatively broad areas of the planet) means that the downlink channel of any satellite internet customer can generally be received over a wide area, even if they are using a modern two-way VSAT. With reasonable encryption this doesn't pose much of a problem, but some satellite ISPs, particularly in the developing world, continue to make use of unencrypted IP-over-MPEG for their downlink. This allows a malicious eavesdropper anywhere in the region to receive the traffic one direction, which has the extremely important implication that it allows them to hijack TCP sessions by intercepting sequence numbers - this enables IP spoofing over TCP, something that is generally regarded as impossible. This is sufficiently reliable that it has been used as part of the C2 infrastructure in botnets. I have a research paper on this somewhere.