Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. I Built Non-Autoregressive Decision Models with RL a Year Ago(convaiinnovations.com ↗)
    159comments
  2. AI-generated posters don’t have to be horrible(john.hartnup.uk ↗)
    522comments
  3. Human brain is two separate organs, Stanford Medicine-led research finds(stanford.edu ↗)
    194comments
  4. Tin: full-text search for Postgres(planetscale.com ↗)
    52comments
  5. A graphical desktop for the ZX Spectrum(github.com/mindbox77 ↗)
    79comments
  6. “The Secret Life of Circuits” is here(coredump.cx ↗)
    58comments
  7. Android 17 is the first since 3.x to add new APIs without releasing to the AOSP(grapheneos.social ↗)
    571comments
  8. Supabase (YC S20) Is Hiring for OrioleDB(supabase.link ↗)
    discuss
  9. Black Holes or Black Hole Stars? Astronomers Spar over 'Little Red Dots'(quantamagazine.org ↗)
    20comments
  10. New evidence for hidden chambers beyond Tutankhamun's tomb(nature.com ↗)
    9comments
  11. GPT-6 Astra Solves a WWI German Radio Cipher(prinzai.com ↗)
    141comments
  12. San Francisco Onion Futures Company(onionfutures.com ↗)
    125comments
  13. Almost Never Use AI to Write Anything Substantive(erichgrunewald.substack.com ↗)
    21comments
  14. What Zig felt like, coming from Rust(besok.github.io ↗)
    134comments
  15. If math is more than proof, we need to better celebrate the rest of it(terrytao.wordpress.com ↗)
    203comments
  16. Cloudflare Quick Tunnels(cloudflare.com ↗)
    301comments
  17. How to Write with an LLM(sockpuppet.org ↗)
    364comments
  18. People who know the most often sound the least certain(vrash.substack.com ↗)
    1comments
  19. Asking Authors About Their Own Papers(medium.com/tmlrorg ↗)
    47comments
  20. Adventures in Microcontroller Circuit Debugging(bigmessowires.com ↗)
    1comments
  21. You can run Git on object storage if you re-make packfiles(tigrisdata.com ↗)
    28comments
  22. Saving another 100TB of RAM(cloudflare.com ↗)
    93comments
  23. Communication by means of modulated Johnson noise(pnas.org ↗)
    21comments
  24. SDCC – Small Device C Compiler(sourceforge.net ↗)
    25comments
  25. Science Is Open Software(jepedersen.dk ↗)
    51comments
  26. OpenJev(openjev.com ↗)
    283comments
  27. Why building a Rust LSP is hard(rust-glancer.github.io ↗)
    50comments
  28. Ray Ozzie and the Optimism of Being Early(reproof.app ↗)
    16comments
  29. The first new cat species discovered in 100 years(nationalgeographic.com ↗)
    128comments
  30. How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip(ieee.org ↗)
    123comments

Dbxfs – A Dropbox-backed file system

111 pointsby 8y agogithub.com
37 comments
8y agoHN ↗

Does this work for mounting Dropbox for Business folders? If so, this solves a significant problem for Linux users of Dropbox (whose official Linux client does not support account switching).

8y agoHN ↗

After Dropbox stopped supporting filesystems besides non-encrypted Ext4 I've dropped it entirely.

I now use SpiderOak One, which doubles as E2E encrypted cloud backup in addition to folder sync.

8y agoHN ↗

SpiderOak looks interesting. The biggest features for me missing from other "sync a folder to the cloud" type services are selective syncing on a machine-by-machine basis and generating public links for my files. These are the only two things keeping me away from iCloud Drive.

8y agoHN ↗

Spideroak does both of those things. They call those features "sync" and "share".

I have a "work" folder that is synced to my work laptop and desktop, but not my personal laptop.

I haven't used the "share" feature, but it exists and it says right at the top of the tab in the client that it's for making public links.

8y agoHN ↗

Just checked them out. They're pricing is super competitive too.

Do they have the 'un-delete' feature that Dropbox has?

8y agoHN ↗

Their main business is personal backup, so they would be remiss to not have a good restore feature.

Each of your computers is associated with a "deleted items bin" where files go when the backup finds you delete them locally. They never get automatically deleted (even when you've reached your backup quota? the website isn't clear).

8y agoHN ↗

They are located in the united states. That alone is more than enough to push me away.

8y agoHN ↗

Personally I was a Spideroak customer for 2 years or so and moved to Tresorit after the canary thing. They offer very similar features but their apps are a lot better. Spideroak always felt quite clunky. Also, the servers are based in Europe.

I don't think either is great to use with lots of data as upload and download seem quite slow, but for my use case of just storing documents it is fine.

8y agoHN ↗

I switched to Syncthing with the main desktop backed up to S3 with zbackup (encrypted). Syncthing works really well (and faster!) when keeping laptops and desktops in sync for my case.

8y agoHN ↗

Just have a service on one of your nodes sync the local directory to S3 occasionally, but SyncThing doesn't do it for you.

8y agoHN ↗

The one killer feature I wish Syncthing had was untrusted peers like Resilio has. There has been an issue[0] open since 2014 but I guess there are still major challenges to this feature.

It would be great for me to be able to specify a certain node as encrypted and read-only like I can in Resilio. If a friend wanted to store his files on my machine and vice-versa, we could do so without being forced to give access to each others plaintext files. Likewise, if I wanted to spin up a VPS and host an untrusted node to help facilitate syncing but didn't want my unencrypted files to be sitting on the disk, I could do so easily.

[0] https://github.com/syncthing/syncthing/issues/109

8y agoHN ↗

can you please explain what scenario this applies....whatis...node/

8y agoHN ↗

How about truecrypt/veracrypt with Dropbox ?

8y agoHN ↗

Nope. I run that (on windows) and got an email telling me I wouldn't be able to use it on that machine any more.

8y agoHN ↗

Yeah, I'm in the same boat. That was a pretty sad move on their part considering their argument about extended attributes made no sense whatsoever.

I personally switched over to MegaSync, I have an account from a while ago that has 50GB space for lifetime, which is enough for my PDFs.

8y agoHN ↗

You can make an Ext4 filesystem in a file, loopback mount that and point Dropbox at it. Works just fine.

8y agoHN ↗

Or just use disk-level encryption (like LUKS).

8y agoHN ↗

SpiderOak has some serious performance issues if you have a lot of data. I just wanted to back up a few hundred GB of data but the client wouldn't do more than ~10Mbit/s on my gigabit line.

Also worth noting their "zero-knowledge" stuff is broken as soon as you log in to the website or use one of the mobile apps, at which point the server has your key.

8y agoHN ↗

We've come full circle from the original dropbox post on here

8y agoHN ↗

See also: https://github.com/archiecobbs/s3backer

Run a normal filesystem on top of S3! (Non-shared.)

In theory it might have really good performance due to your kernel caching blocks and files, and 25Gbit throughput to S3. Dependent of course on your instance being in EC2 and in the right region and having a big enough instance to get 25Gbit network.

I tried AWS EFS and found the performance very sad. Like 100Mbit even with a 25Gbit instance and the highest-specced EFS filesystem.

8y agoHN ↗

EFS performance was tied to how much storage you were using (at the rate of 50KiB/s per GB stored… with the ability to burst to 100MiB/s based on a credit system. A file system storing 256GiB can sustain 12.5MiB/s and burst to 100MiB/s for up to 180 minutes per day.

If you like, you can now specifically provision throughput for EFS at the rate of about $6 per MiB/s (8Mbps) per month.

8y agoHN ↗

I provisioned throughput on the EFS filesystem to the largest allowed value and still got crappy performance. Maybe it was the access pattern of the data, but large bulk file transfer performance was no higher than 100Mbit on a 25Gbit instance...

8y agoHN ↗

check this one out: www.lucidlink.com

more than s3 sync

8y agoHN ↗

Waiting for someone to implement dropbox-based deployment of a PHP site.

8y agoHN ↗

In light of what Storj.io is about to bring to market, I'm curious to see if this could be modified to work with it. I'm a huge fan of Dropbox and have used it almost every day for the past 7 years, but if a decentralized option comes along with feature parity and comparable pricing, I'd switch in a heartbeat.

8y agoHN ↗

The author should think about applying to Dropbox ;)

8y agoHN ↗

Isn't the author a Dropbox employee? Not sure if he's still at Dropbox but I believe he's one of the first employees.

8y agoHN ↗

That was the joke, newhouseb used to be in charge of the sync team at Dropbox :)

8y agoHN ↗

I'll just note that you can mount Dropbox as a file system with rclone ( https://rclone.org/ ) using the rclone mount subcommand.

rclone mount also works with the 20 or so cloud providers supported by rclone (s3, gcs, swift, box, dropbox, b2, etc).

rclone runs on Linux, Windows, macOS and quite a few other architectures! It is written in Go.

8y agoHN ↗

Fwiw Dbxfs seems to be a lot faster than rclone