Post #1225014

2026-04-04 22:16 UTC

I think TCP window scaling is supposed to deal with that. But I have also heard that computers used on ISS supposedly all do remote destktop to computers on the ground when they need any outside communication. But I don’t recall the source, so I am not sure if this is accurate. Speaking of 1+ sec latency, some point in the past year I discovered that BIND 9 would completely fail to perform recursive lookups once latency exceeded 1 second. That came as a surprise to me as prior to that my experience had been that it was a robust DNS implementation.

Replies (3)

@revk@toot.me.uk 2026-04-05 06:28

@kasperd @jamesthomson TCP spoofed ACKs and a more robust transmission over the slow medium is a good trick for such things. But applications timeouts may also need tweaks. DNS is a good one, but a suitable DNS cache which has long timeouts can help. Similar to satellite comms issues. I assume they have some clever stuff! Remote Desktop must be frustrating.

Open ##1225015
@grahamadams@toot.wales 2026-04-05 07:35

@kasperd @jamesthomson @revk Perhaps it was falling back to TCP and DNS over TCP was blocked by the network? I have come across this before.

Open ##1225024
@sborrill@justfollow.me.uk 2026-04-05 07:39

@kasperd @jamesthomson @revk On the ISS, it was #Citrix XenDesktop for the desktop access and Branch Repeater to help with latency (uses lots of highly desktop-optimised caching). These are both the old product names. https://www.informationweek.com/it-leadership/nasa-taps-citrix-for-tweets-from-space

Open ##1225037