Hacker News Reader

Comment by Retric | Hacker News Reader

Anyone have more information about why this is needed?

There are already decades worth of ground to satellite network connections in use, and satellite to satellite communication is old news. Past astronauts have also used internet connections so what problem does this solve?

ed: Or is this just a test for future connections?

View on HN

This paragraph seems to explain why:

"DTN works by providing a reliable and automatic “store and forward” data network that stores partial bundles of data in nodes along a communication path until the parts can be forwarded or retransmitted[...]. This differs from traditional Internet Protocols that require all nodes in the transmission path to be available during the same time frame for successful data transmission."

I think there are plenty of situations where end to end communication links are temporarily down because something is on the wrong side of a planet or moon, having things held for retransmission as close as possible to the final destination would improve on this.

Isn't this just normal packet networking but with big packets and no timeouts?

Store-and-forward is an idea at least as old as Usenet. This also reminds me of the way that `%ames` networking in Urbit works. There is more to it than you say, you may want pub-sub from the beginning and features like exponential back-off, without a source to cite I'd say I think intuitively some of these things are likely to be missing from regular packet networking implementations.

No, because routers do not retransmit packets, only hosts.

if you had something like

earth -> sat -> sat -> mars

and the equipment on mars is temporarily unreachable, you don't want to have to repeatedly retransmit

earth -> sat -> sat

since that takes between 4 and 24 minutes end-to-end. You want the data to store on the remote sat until it can be forwarded to mars.

That switch better have one helluva buffer.

This probably means they should adjust their browser and DNS timeouts to something like half the time to complete an orbit.

This is for future use . . . Imagine an "internet of things" where the things are space probes spread throughout the solar system, landers on other planets, and so on. Because of light speed delays, as well as availability issues due to the satellite being in the other side of the planet, for example, new protocols and techniques are required to ensure reliable communications.

testing "store and forward" disruption/extreme long latency networking stack technologies and software.

it's only 495ms round trip ping time from earth via a geostationary orbit telecom satellite to another location on earth. geostationary works fine for local stuff and works with TCP/IP.

latency to mars is between 4 minutes and 24 minutes depending on the alignment of the planets. big difference.

I wonder if this is an extension to BGP for each router to know when to transmit and in what direction.

My guess is that each node decides that locally, since it can't rely on a constant flow of information from the network. Also there's a matter of power optimisation.

I'd love to see a more in-depth description of this, perhaps they're going to publish a RFC:)

They've published multiple RFCs.

https://sites.google.com/site/dtnresgroup/home/docs#TOC-RFCs...

I'm no expert, but I believe this was created to deal with the massive delay in communications between Earth and Mars. The ISS is in low Earth orbit, so it has less than a second of delay in radio signals. Mars is 20 light minutes away, so there's a 20 minute delay between with all packets sent.

My guess is that astronautical internet to the space station doesn't have high enough latency to cause problems, so 'normal' protocols can be used happily enough.

But if you want to link a colony on Mars to Earth's internet, you'll need something like this 'store and forward' style networking, rather than trying to rely on interactive two-way realtime-ish communication.

Reply on news.ycombinator.com