Maybe a big ask here, any thoughts on the backburner ATM about extending the plumbing with a Syncthing-like capability?
As always, I’ve got reckons about the efficacy of different models for maintaining a single copy of the truth, plus backups of course.
SAN, NAS, Cloudy thingy, with the network latency and all the possible weird protocols to edit the single copy on a local device, or eventual consistency with a background synch task between the remote and local filesystems.
Seems to me that Tailscale’s mesh optimisation of network paths enables, in many common situations, a reasonable tradeoff between latency and userland understanding of what is being edited on their local device.
Sure it’ll never be PCIe fast if synching over most cellular networks is involved, and a large MP4 will need regular RTFM prompts for some … most users.
But isn’t there a huge synchronicity here between Tailscale’s infrastructure and my darling, pick two of the three CAP theory?
Whoa there Sōf, brakes on the squirrelbrain.
Hey, was I suggesting TrueNAS integration too … but now that you mention it.
All sorts of backburner thoughts, sure. Brad Fitzpatrick started and is a main developer of https://perkeep.org/, for example, so there is a familiarity with the space.
Regarding TrueNAS, Tailscale does work on FreeBSD and I’d expect it to work on a TrueNAS system.
Lots of neat experiments are possible nowadays on top of tailscale, but we want to generally minimize the number of things that end up directly inside the tailscale app unless they’ve proven to be really useful.
Taildrop was a bit of a special case: we wanted to try making a user-friendly, but technologically very lightweight, feature. Something like syncthing is more technologically advanced, and harder to put a nice UI on top of. So I might suggest experimenting with just running syncthing yourself and letting it run over tailscale. If there are specific integration points that would help, we can talk about adding them if needed.
(Specific example: we introduced the new “peerapi” for finding tailscale peers with certain capabilities. The plan is to expand that to be more generally useful. For example, maybe your syncthing nodes could find each others’ IP addresses through there. Or maybe we should do something with mDNS discovery…)
Wheeeee, more more things I probably don’t have time for but are fun to play with at 3:00am when I can’t sleep.
Taildrop as a special case without a big picture driver makes makes complete sense, intuitively. It’s a basic plumbing thing.
It was watching the inevitable IM app cycle of doing text, then text files, then voice calling, then video calling, then huge video files … but no email client, so sad, but so sensible… combined with my predilection for, as I said previously, single copy of the truth files, that prompted my rather inane question.
So, many thanks for taking the time to reply, it’s always good to get glimpses of where a product’s heading, even if there’s sudden course corrections occassionally.