Alpine also has the advantage of musl, which is a safer alternative to glibc, at the cost of some performance. So, if anything, I’d expect people to consider alternatives to Alpine for that reason, as alpine is already the best choice for security.

good idea, but a slight correction - mDNS and SSDP are entirely different things, rather SSDP was Microsoft’s initial proprietary take on the idea, and mDNS was created as the new and improved standard.

mDNS does multicast (send/receive) on 224.0.0.251 and ff02::fb on port 5353, while SSDP uses 239.255.255.250 on port 1900 as you mentioned.

But I think OP’s issue is that they’re on different subnets; mDNS expects the server and the client to have a perfectly overlapping subnets inside the same LAN. If the server has 10.0.0.3/16 and 192.168.1.3/24 then the client must also have 10.0.0.7/16 and 192.168.1.7/24. Or, if you can tell the server software exactly which IPs to announce, then that might work too.

there are workarounds to this, using avahi reflector, but that thing is buggy – specifically you need to disable NSEC on the server, and lock it to either IPv4 or IPv6.

I made some note on additional pitfalls while i was writing my own mDNS and SSDP servers for fun, they’re at the bottom of this page: https://ocv.me/copyparty/helptext.html

Absolutely; if I was a company, or hosting something important, or something that was intended for the general public, then I’d agree.

But I’m just an idiot hosting whimsical stuff from my basement, and 99% of it is only of interest for my friends. I know ~everyone in my target audience, and I know that none of them use a VPN for general-purpose browsing.

As it is, I don’t mind keeping the door open to the general public, but nothing of value will be lost if I need to pull the plug on some more ASN’s to preserve my bandwidth. For example when a guy hopping through a VPN in Sweden decides to download the same zip file thousands of times, wasting terabytes of traffic over a few hours (this happened a week ago).

and filtering malicious traffic is more important to me than you visiting my services, so I guess that makes us even :-)

I’ve started blocking entire AS from my servers; since I only expect/want traffic from home ISPs it’s kinda alright, but still… man

HFSv2 (the windows exe) has publically known unfixed vulnerabilities! Please upgrade to HFSv3 (nodejs, crossplatform). I’m also maintaining a list of other alternatives; https://github.com/9001/copyparty/blob/hovudstraum/docs/versus.md

You could always toss it in a sandbox for some isolation :> but yeah I get you, all of the optional features does mean more code.

It’s a shame that browsers make stuff like chunked uploading so tricky, so even just the essentials would be a fair bit of logic – and you won’t get optimal upload speeds without sending chunks in parallel. And the corruption detection is also worth its weight in gold… Ah well, it is what it is hehe

Thanks for the mention :>

Yeah, copyparty was my attempt at solving this issue - a single python-file for receiving uploads of infinitely large files, usually much faster than other alternatives (ftp, sftp, nextcloud, etc.) especially when the physical distance to the uploader is large (hairy routing).

I’m not gonna put an upload on my site, that’s a security nightmare waiting to happen.

curious to hear your specific concerns on this; maybe it’s something that’s already handled?

You may encounter some confusing errors with MergerFS; depending on your config, any software using SQLite may crash with a scary-sounding “disk I/O error”. The quickstart has more info, https://trapexit.github.io/mergerfs/quickstart/

for a selfhosted service which is a single self-contained process in a single container, is there still a benefit to using compose, and if so, what would that be? genuine question since I’m not providing a compose example for a foss service I made.

Awesome!! And thanks for mentioning the traefik thing; I added it to the example so maybe it’ll help somebody else one day :>

hey, copyparty dev here - I’ll be upfront and warn that if you’re looking for a service to do full bidirectional file syncing, with proper file tracking and versioning, then something like syncthing is a better choice. Copyparty is able to do single-direction syncing of local folders to the server pretty well (using copyparty’s u2c or rclone), but that’s about it.

But if you’re looking for something to handle file uploads faster than many alternatives, or any of the other features listed in the readme, then I’d be happy to help if you ever get stuck somewhere :>

Like others have mentioned, I wouldn’t trust the iis WebDAV server any further than I can throw it. And moreover, the WebDAV client that’s built into windows is also good for nothing – it has a filesize limit because it reads the whole file into ram, instead of using http206 like any sane server/client. And it also has a chance of crashing explorer.exe after reading a couple thousand files…

That’s why I’ve been making my own WebDAV server, but I’m also keeping track of other alternatives. And for connecting to it from windows I’m using rclone. Regardless which server you choose (just please do not use iis lol) you can borrow these examples for connecting to it :-)

I’ve been making copyparty, and one of the initial inspirations/usecases was exactly this one – bunch of internet-friends meeting in meatspace for some event, and needing to stream across pics/vids before we split up and maybe never meet again :-)

You can run the server on android using termux; see the instructions in the readme for the installation part, and then run the server like so: copyparty -e2ds --qr -v /storage/emulated/0/Pictures/from-family/::w

it will show a QR-code which you can scan from the other device to start uploading, and the folder is shared write-only (:w instead of :rw), so nobody can download anything. And it keeps track of which files have already been transferred, so the client will skip across dupes without reuploading them, and also resume interrupted uploads automatically.

Tried sending 340 photos (1.21 GiB) from an old iPhone to an android hotspot just now; I turned off the iPhone screen as soon as it started sending, and it finished in 41 seconds, so around 30 MiB/s? Maybe there are faster alternatives… But my javascript should be fairly resilient and recover from network glitches and such, so uploading from safari like this should be fine.

Not proxmox-specific, but I’ve been using btrfs on my servers and laptops for the past 6 years with zero issues. The only times it’s bugged out is due to bad hardware, and having the filesystem shouting at me to make me aware of that was fantastic.

The only place I don’t use zfs is for my nas data drives (since I want raidz2, and btrfs raid5 is hella shady) but the nas rootfs is btrfs.