Hmm. If storage continues to improve as fast as it has been, once the space infrastructure is there, this might be cost effective. We have a lot of data to send.
In particular, we are going to need to have a local cache of the internet on Mars. At best, the average minimum latency is 225 million kilometers / the speed of light = 12.51 minutes.
Vast data barges, roaming the solar system, keeping everyone in sync....
Building an interplanetary Internet is going to be very, very fun.
On a related note, never underestimate the bandwidth of a truck filled with...
Well I was going to say blueray disks. But the cheapest I can find bluray disks are 3 cents per gigabyte. 3 TB harddrives are easily had at 5 cents per gigabyte. If the data only needs to go 60 miles, then 120 drives ($18,000) in a truck would get 1 TB/sec, or 8,000 mbps...
Ouch. I'll go with a megabit connection. Once upon a time that sort of thing worked for bulk data ;)
I predict somewhere today at Google an engineer just got an idea for his 20% project. "Red Cache"
Straight snapshots would probably be the only viable option though. And maybe make the policy, once a decade do a complete update. And every year, do "any webpage accessed by colonists in the past year will be updated. And the top 5-15% of webpages by popularity on Earth."
And during the year, the top 5% of webpages, and any webpage with specifically Mars relevant content would be updated "wirelessly."
The numbers would change constantly, depending on bandwidth and costs and whatnot. But the question "how to build the best asynchronous shadow Internet given such and such constraints" is fascinating.
-- A message from your friendly local hacker wizard
0 latency on funny cat videos from Uranus.
It’s not like that rover is driving around autonomously. It has some autonomy, but it can’t decide what’s interesting and what’s not, what has to be investigated and what not.
This means if you're already in Mars orbit you need fuel(TM) to escape. One might consider a slingshot around Deimos or Phobos but I don't think that'd be helpful enough to justify its cost (could be wrong, but I've never seen it suggested in any Mars->Earth mission plan).
(assuming 8-bit bytes)
But in a pedantic sort of way, the computation should add extra bits for ECC since the drive has ECC bits already embedded in it, so a 1TB drive probably has more than 10 Tbits it is probably closer to 15 - 20 terabits.