Hacker News Reader

Comment by anthk | Hacker News Reader

Back in the day there was an open officialish IRC channel from the Antartica researchers.

Sadly I missed that.

Also, on the limited connection they have, mosh and IRC among mail/nntp could be life savers for comms as they can be slow and unrealiable. If they work under mosh against a public Unix server they could discuss nearly anything.

On images, Imagemagick can do dithering magick and save lots of bytes under a JPG/PNG format with 256 colours.

Also, if the goal is to send data even with connection cuts, nncp can resume data packets on the fly until it finishes.

View on HN

I routinely ssh into my experiment at the south pole during satellite windows. I just use tmux on the other end...

How is the latency as you type?

Same as any non-LEO satellite link (probably like, O(700 ms) latency? About the same as our experiments in Greenland...). Not great, but fine for administration and light vimming. The worst was when I had to update some py2 to py3 code after a system was updated, and there was no way to test except on the South Pole machine. The places where bytes vs. strings bite you when dealing with serial devices are plentiful...

This stuff is always so fascinating to me. I remember I was on a plane once and I had done a code deploy before getting on the flight, then remembered I forgot to upload the necessary files to the k8s cluster. There's no way I had enough bandwidth on a plane to get the files on the cluster. So I tried ssh-ing to my home machine and doing it from there -- nope, SSH is blocked on the stupid in-flight Wifi. So I went on DigitalOcean, got a $4 instance, and used the web interface to SSH into my home machine, and then from the home machine uploaded the files to the cluster. The latency was hilariously slow, I think over 1 second, but it worked!

I was thinking about the route those packets were taking - from my computer to the plane router, up to a satellite, down to the DO data center in NY, from there to my home in NJ, then to the K8s cluster in NY, and then all the way back around. When you put it that way, one second per character isn't that bad.

With mosh you can even resume connections after disconnecting but without losing interaction with the target machine as it happens under SSH. Also, it will use far less bandwidth. I tried it with 2G/ISDN-DSL like speeds and it worked great. And you can keep using tmux just in case.

On the "caveats", you need UTF8 support in the client machine, but today that's granted.

I'm sure there are some advantages, but I couldn't quickly figure out how to get it running with the various jump hosts required (and I also often use port forwarding...). Just having tmux on the other side gets me most of the advantages, I think...

Well, on port fwd, maybe if you forward the mosh server port would work. If you use SSH itself as a tunnel, the setup it's far complex, as mosh is just a remote shell.

Still, mosh curb-stomps ssh on latency, input guesssing, data saving and bandwith usage.

Using that over 2G was a literal black/white difference compared to SSH which was nearly impossible to connect and the latency was unbearable. Under mosh I could chat, email, nntp (they still are some good newsgroups), IM over jabber/telegram, read news over RSS, code in C/TCL (jimsh) and read the docs with links, among reading books (novels and essays) with some CLI epub reading script.

Yes, I also used tmux on the remote machine too, mosh+tmux work great with each other. Albeit mosh resumes a potentially lost/dropped connection just fine (even without a running remote tmux/screen), I think tmux itself would send less data over the wire.

Can you talk about this a little more?

Sure... I work on experiments where we basically we instrument ice sheets with radio antennas and look for radio emission that might come from really energetic particles smashing into the ice. One of the experiments I work on (the Askaryan Radio Array) is at the South Pole. Each detector has a single board computer to take data and such. They are cabled via fiber to the IceCube lab at Pole, where data is relayed to. About half the time, the South Pole has internet either via NASA satellites or DSCS (there is also Iridium all the time, but that is not for general use), and via some jump hosts, I can ssh into the detectors.

Very cool, Thanks!

Their website (in bio) looks relevant.

Is starlink available?

It's a delicate question.

Starlink does have some satellites in polar orbit (initially it seemed like there wouldn't be any, so this is relatively new), and there were some limited Starlink tests last year, but due to the CMB telescopes at Pole, which may be negatively impacted both by the satellites and the ground antennas, any Starlink deployment will be very cautious.

Tested and rejected due to interference

Studying coexistence is an active area of effort (we even applied for some NSF funding to study this...)

Is that still true? With the size of the bases, I had assumed the network connection would be significantly more robust and/or higher bandwidth. How does the Icecube project or similar scientific experiments work? Or is it all still station wagons full of tapes?

If by "station wagon" you mean LC-130 ;) IceCube switched from tapes to hard drives I think in 2015 or 16.

In 2014 we spent a lot of time fiddling with tapes and drives... At that point, all the data was recorded to tape and flown North in the summer. Additionally, software at Pole filtered out the more interesting events to upload over satellite within hours. I assume it's still basically the same.

I'm curious about whether the temperature at the pole presents any difficulty for handling tapes (and hard drives). How easy is it to keep media within a safe temperature window in situ, and during transport in/out?

Directly, the temperature isn't as big an issue for operating tapes/drives/computers, since they are inside temperature-controlled buildings. There are indirect problems though; the incredibly low relative humidity that results from heating outside air - that dryness leads to problems with static. Various solutions were tried to keep the tapes humidified, but in the end I think everyone was relieved to switch to hard drives :). South Pole is also fairly high elevation, which combined with the low humidity makes air cooling work a lot less well than it might in other places - the air just doesn't have as much heat capacity as in more moderate environments.

As far as transport, it's about a km (in winter, typically a walk) from the main station to the "dark sector" where the CMB telescopes and IceCube Lab are. As another comment points out, there are cables between the main station and the dark sector so typically data flows through them. When you do need to move equipment, it's usually not too big of a deal to carry it inside a parka or something if it's smallish, or arrange a vehicle for bigger items.

Things like cables can be a bit of a learning experience; a friend hand carried a PVC-insulated monitor(?) cable outside, coiled up, then promptly tried to uncoil it and found that PVC gets quite brittle at low temperatures...

Reply on news.ycombinator.com