SoylentNews is people
SoylentNews
Boffins [Scientists] that want to see Internet protocols extend to outer space – the so-called “Interplanetary Internet” – need to prove they're offering something useful, according to one of the father-figures of the Earth-bound Internet.
Vint Cerf, who has taken an interest in beyond-Earth applications for the Internet protocol stack since the 1990s, told last week's InterPlanetary Networking SIG (IPNSIG) meeting that to get beyond a mere curiosity, the SIG needs to be useful.
“Our challenge, to the extent that we're interested in serious expansion of communications capability for space exploration, is to demonstrate its utility,” Cerf told the gathering.
“It's not that anyone thinks that you should just build this interplanetary thing and hope that somebody uses it,” he added.
One possibility, for example, is that spacecraft that support these kinds of protocols could, having fulfilled their primary mission, have a longer economically-useful life if they can then become nodes in the interplanetary backbone.
And there's no doubt that there'll be a lot more data being flung around in space: last year, for example, the success of NASA's LADEE broadband experiment showed that free space optics could cook along at hundreds of megabits a second without an atmosphere to get in the way.
http://www.theregister.co.uk/2015/05/27/interplanetary_network_sig/
IPNSIG presentations and videos: http://ipnsig.org/2015/05/26/speaker-presentation-materials-2015/
Original Submission
(Score: 4, Informative) by VortexCortex on Saturday May 30 2015, @05:53PM
I've implemented a modified version of NASA's DTN (Disruption Tolerant Networking). [nasa.gov] It's great for terrestrial use. If the FCC and powers that be would have gave the public back an unregulated slice of the airwaves they got back during the switch to digital over the air TV, then we could implement a mesh network using DTN (space Internet). It's decentralized, and while designed to work via "line of sight" connections, the system also lends itself to the one-to-many properties of HAM packet-radio. Store-and-forward is great. It means that the data moves closer to where it's needed. A tiered caching network by default = free collocation.
NASA's version is a bit flawed, they're still getting down packet level implementation details, but the higher level proposals I've seen still have deduplication via URL (asset name), ignoring that "everything is a file" is just dumb (literally ignorant), everything has a hash is a better approach and avoids the namespace collision issue. The name of data doesn't matter, that's just arbitrary metadata, POSIX allows certain paths to have special meaning to the OS, but when it comes to data itself thinking in terms of names is literally retarding. Instead I use a hash of the data to reference the data, a human readable named link is just an alias resolving to an infohash, so that way if multiple "directories" link to the same data by different names they'll still have the same set of infohashes for the deduplicated data chunks. Separate form from function (separate symbolics from realities), ergo separate content from style and separate names from data. There are some filesystems that already do something similar.
Implement the Interplanetary Internet protocol here on Earth. Instead of funneling traffic into a pipe you can pull your data from your neighbor who just watched the same thing, or from the next cache up the pipe. A whole planet of people need only transmit that cute cat video once across the interplanetary link instead of maintaining billions of active streams to the source, and with infohashes you can avoid multiple renamed copies being sent. With chunks instead of files you can even share clips or excerpts of a video without retransmission of the asset itself. Latency is only an issue for interactive resources (like IRC servers), but using a bulletin board + "push notification" works around that. The key is that most nodes will be receiving additional data rather than what they requested themselves in order to facilitate the store-and-forward. Instead of imagining the network as endpoints that serve requests think of the network as a whole propagating a swarm of data across it. This causes some waste in that some cache data may never need to be retransmitted, but that's simply the cost of redundancy and reliability.
The FCC has strict regulations against packet radio, even on the family band. This is because such a "space-age" decentralized distributed deduplicated network is exponentially harder to spy on -- You have to put up listening posts between every node instead of just taping a trunk line: Am I reading a site or getting it for my neighbor? After I have some data, how many of my other neighbors could have downloaded the data from me? This is much different than logging all requests to the nameservers... Additionally, a mesh node need not pay subscription fees to the mesh other than its cache and relay facilities, thus you buy your gear once and only pay for maintenance instead of exorbitant fees for connectivity and bandwidth caps. DNT mesh would threated the existing cellular network's profit margins, and a proper hash-based deduplication system would greatly reduce ability to spoof data. And that's why we aren't already enjoying the benefits of a DTN here on Earth.
That said, implementing an Interplanetary Internet is far more costly than a DTN mesh here on Earth -- I think that's what Vint Cerf is arguing against, and why I'm arguing for a terrestrial trial run. The thing is, we already have space probes out there, and we'll only have more in the future. Implementing a decentralized deduplicated distributed protocol that turns all nodes into relays and caches and allows us to route around the sun and not maintain direct links at all times with all probes is the ONLY logical thing to do. Eventually we'll have multiple probes that require the same data (perhaps a firmware update), they should be able to update from each other. My implementation started out as a way to load balance internal OS update traffic. The question is not whether the Interplanetary Internet should be built, it's when will we be forced to do so? It's already proven useful, and due to the economics of demand and relativity DNT is the inevitable future of both interplanetary and terrestrial Internet will adopt. As our computing platforms get more complex they'll also have such distributed systems (see: Supercomputer clusters). A DTN is the answer to light's limitations, not Vint Cerf.
I'm sorry, if you lack the foresight to avoid the foolishness of running a centralized "web" atop a decentralized Internet by not building caching and deduplication into the network (ignoring store and forward best practices), then I don't particularly value your opinion of the solutions to the massive problem you've created.
(Score: 4, Interesting) by kaszz on Saturday May 30 2015, @06:19PM
What are the specific FCC restrictions on packet radio on the family band etc that hinders DNT? I was under the impression that packet radio is common among radio amateurs? (albeit slow at 1200 bit/s or so).
Perhaps this limitation can be circumvented by laser and other less restricted means of transmission?
Wouldn't your hashes in essence be the name space? and how would low probability but still possible collisions be handled?
Doesn't T.o.r and I.2.P have large similarities to DNT?
Any good papers on DNT?
(Score: 1) by nethead on Sunday May 31 2015, @01:23AM
Most hams are using 2.4GHz WiFi for mesh networks. The lower part of that band is a ham allocation.
https://www.youtube.com/watch?v=pryc8jIl6Xo [youtube.com]
How did my SN UID end up over 3 times my /. UID?