NASA space tests 'interplanetary internet' protocol

Mailto: glop_the_conqueror@mars.planet

"The Interplanetary Internet is designed to be more robust, so as to cope with the huge latencies of space communication"

So a game of counterstrike with a Martian is out then? The ping's a bitch.

Still a Tracrt might be interesting.

I claim the Domain name

Google Mars

@ Chris G

Pity .et has already gone

Why not......

....use subspace transmission. It would avoid all the long delays.

Delay tollerence

TCP/IP has timeouts in it to allow it to detect dead connections. In fact the linus tcp code actually has the following comment in it highlighting the problems of talking to mars:

From tcp_timer.c:

/* Increase the timeout each time we retransmit. Note that

* we do not increase the rtt estimate. rto is initialized

* from rtt, but increases here. Jacobson (SIGCOMM 88) suggests

* that doubling rto each time is the least we can get away with.

* In KA9Q, Karn uses this for the first few times, and then

* goes to quadratic. netBSD doubles, but only goes up to *64,

* and clamps at 1 to 64 sec afterwards. Note that 120 sec is

* defined in the protocol as the maximum possible RTT. I guess

* we'll have to use something other than TCP to talk to the

* University of Mars.

*

* PAWS allows us longer timeouts and large windows, so once

* implemented ftp to mars will work nicely. We will have to fix

* the 120 second clamps though!

*/

@ Sian Williams on the ball

That's the BBC for you. A few years back the Today programme did an item a about the NASA Mmercury probe.

John Humphrys: "But whats the point of bringing it back?"

Baffled NASA spokeswoman: "We're not!"

John Humphrys: "Oh. I seem to have been misinformed.""

And there was the wonderful Digiguide review of a science programme that started "Despite being closer to the Sun, Mars is colder than the Earth because....." Later "corrected" to "Even though it is further from the Sun, Mars is colder than the Earth because....."

The NeXXXXTSTEP for Jobs?

"Anyway, I thought TCP/IP was delay-tolerant to a degree?" .... By Anonymous Coward Posted Wednesday 19th November 2008 12:07 GMT

I think by design, that TCP/IP is delay-inherent ... for Spin Waiting/Vetting purposes .... but that only works if one waits on/for Spin rather than being more Content with Generating it for IT and Media.

"Sounds like someone's trying to get funding for "The internet, right, but -get this... IN SPACE"" ... Crikey, AC, that would be right in the Heart of Alien Territory.

New technology

"Unlike TCP/IP, the DTN does not assume a continuous end-to-end connection. If a destination path cannot be found, data packets are not discarded. Each node keeps the information as long as necessary until it can communicate safely with another node."

Ah, that'll be uucp then. Might need to crank the window sizes up a bit though.

@AC: "I thought TCP/IP was delay-tolerant to a degree?"

Sure - it buffers all data at the sender until acknowledged by the receiver. But you still need delay*bandwidth worth of storage. The TCP Window Scale option lets this be up to 1 gigabyte (8 gigabits), giving you ~7Mbps to Mars with a 20-minute RTT.

But if the application decides that the connection has dropped you'll lose it all.

$ ftp downloads.mars.net

<< 20 minutes later >>

Timed out waiting for login

Why not TCP?

I dare say TCP is delay tolerant, but Mars is (depending on the relative positions in their orbits) typically about an hour away from earth if memory serves. That's two hours waiting for a reply, by which time you've not sent a terabyte or two of data. Even with jumbo packets, that's one hell of a sliding window you've got there and Mars is *close*.

And yes, the principles involved in working around this problem are not new. UUCP is an example, NNTP and SMTP are others. (Multiple MX records, for example, solve the problem that the ultimate recipient might actually be *down*, not distant, and push the data onto a near neighbour instead.) So there's nothing new here, unless you count the engineering task of making it work. But the same could have been said for TCP/IP at version 4, and I think most informed observers reckon Cerf and co did a pretty fine job *there*. :)

Jumbo Buffers

I agree TCP is not the right medium for interplanetary comms, but I don't agree that it is purely on the basis of the massive buffer required for stable communication, because that is exactly what would be required for the DTP. The only way to ensure that a packet has gone through, is to keep a copy of the packet, and wait until you've had a response from the receiver notifying you they've received that packet. Either in TCP or DTP, that still means holding onto that packet (and the millions of others that you'll be waiting a very VERY long time for).