> So there is a cross over point where the latency of a LEO satellite system is actually superior despite the uplink/downlink path.
There is, but given the rise of the use of edge caching and edge compute for latency-sensitive applications, with edge locations located to optimize access to masses of users, that crossover is unlikely to be actually reached in many of the cases where latency is actually a concern, and Starlink is going to be stuck behind unless it is able to both get edge caching and edge compute on its satellites and get content/service providers to use it, and comsats aren't exactly optimal locations for that.
Latency is also much less important with highly cacheable content. At least, I can’t really think of examples to the contrary. My patience for loading pages browsing is many orders of magnitude higher than realtime uses. Bandwidth seems to dominate in this context.
Latency of multiple seconds, like you can see in some places around the world, is much more frustrating than the kind of latency discussed wrt Starlink.
The lack of a close edge node is kind of interesting, though, so i’m not trying to dismiss your comment in the least.
There is, but given the rise of the use of edge caching and edge compute for latency-sensitive applications, with edge locations located to optimize access to masses of users, that crossover is unlikely to be actually reached in many of the cases where latency is actually a concern, and Starlink is going to be stuck behind unless it is able to both get edge caching and edge compute on its satellites and get content/service providers to use it, and comsats aren't exactly optimal locations for that.