I'm migrating from a crypted gdrive remote to a crypter dropbox remote.
Files correctly uploaded to gdrive gives file name too long when syncing to dropbox.
how can I avoid this problem?
Run the command 'rclone version' and share the full output of the command.
rclone v1.62.2
os/version: Microsoft Windows 11 Pro 22H2 (64 bit)
Line 4274: 2023/06/01 10:40:18 ERROR : Da sistemare-Rinominare/CARTONI - C'era una volta Pollon (1982)/1ok/C'era una volta Pollon - 01 - La figlia di Apollo (1982 - WEBmux x264 1080p - Ita Jap AC3 - Softsub Ita) [by TAERR Maleno85 Kagome & Jagger T7ST] (1).mkv: Failed to copy: file name too long
Line 4275: 2023/06/01 10:40:18 ERROR : Da sistemare-Rinominare/CARTONI - C'era una volta Pollon (1982)/1ok/C'era una volta Pollon - 04 - La rivincita di Narciso (1982 - WEBmux x264 1080p - Ita Jap AC3 - Softsub Ita) [by TAERR Maleno85 Kagome & Jagger T7ST] (1).mkv: Failed to copy: file name too long
Line 4276: 2023/06/01 10:40:18 ERROR : Da sistemare-Rinominare/CARTONI - C'era una volta Pollon (1982)/1ok/C'era una volta Pollon - 05 - L'arco di Eros (1982 - WEBmux x264 1080p - Ita Jap AC3 - Softsub Ita) [by TAERR Maleno85 Kagome & Jagger T7ST] (1).mkv: Failed to copy: file name too long
Please note that Dropbox has max path limit 260 characters - using encryption makes it even less. base32768 will make it less painful than default base32 but with path like in your example:
Da sistemare-Rinominare/CARTONI - C'era una volta Pollon (1982)/1ok/C'era una volta Pollon - 01 - La figlia di Apollo (1982 - WEBmux x264 1080p - Ita Jap AC3 - Softsub Ita) [by TAERR Maleno85 Kagome & Jagger T7ST] (1).mkv
it can still be a problem. It is 222 characters + 5 characters for Media folder = 227 characters
Maybe it will work - try - but with Dropbox limitations I would recommend to try keep names shorter.
I was going to do some maths to work this out, but then I remembered rclone can do this!
This tests the max file length for a base32768 encoded crypt running on dropbox
$ rclone test info -vv --check-length dropboxsecret:file-length
// dropboxsecret
maxFileLength = 463 // for 1 byte unicode characters
maxFileLength = 231 // for 2 byte unicode characters
maxFileLength = 154 // for 3 byte unicode characters
maxFileLength = 115 // for 4 byte unicode characters
So provided you don't have too many non-ASCII characters you should be able to fit 255 char long file names no problem.
The above is for maximum size of a single path unit.
Despite the docs I don't find any evidence that total path lengths need to be kept under 260 characters. This 508 character path works fine. The individual path elements do need to be kept under 255 unicode characters though.
I think that 260 number is from Windows (without using UNC paths) and dropbox is just being conservative here assuming you might look at your dropbox on Windows and some other platform.
Thx both for the reply.
the base32768 remote test say this for example: 2023/06/01 17:58:51 DEBUG : 姜頲ꌧ匧榴⤗✒䂇恟/盼畉钓灙㨴ꌐꍛ䤿䍷㥷又跕ꞙ羞鶂弚彡夏桹耿裼軖㪬赏㞋腓繻螆䠋Ꝕ愨帰紗阠釭禖ⵌ鋷㴇阓矒嬊吂䌄濇䅼蟚牻槰嫚毪蝥玪隅憥剽Ꙙꏈ筦䋈坱蓗磟樨鑯邦ꍾ餈ꂗ挦汫䯧曰幤剒荨圓㚅㫡巍虻醍郣蛝㔙㴘殰鍶慘闌鴮㇋怔洧: Uploading chunk 3/6
is correctly configurated?
Quick question, could using the base32768 encoding instead of the default encoding cause any problems when retrieving/decrypting the files again?
I've already copied over 80% of my stuff to dropbox so changing the encoding now will upload the rest of the files with the new encoding but what about the other files uploaded previously? Will they need to be retrieved without the filename_encoding = base32768 line in the config?
You need to create a new clone remote and add it the base32768.
After that you have to move all the files from old to new so rclone re-encrypt the files/folders with the new base..
No - unlike base32 (which is 100% safe) its implementation contains case sensitive characters (199 out of 32768)
But chances of a real problem are quite small. Based on ncw maths from linked thread:
I also tried harder to calculate the probability of this happening, and I think you'd have to put around 10²¹ files in a single directory before you got a collision, so I don't think it is very likely to happen ever.
The potential issue was raised with base32768 author:
but fix wont come from there.
Solution is to create new "base32768" implementation with case insensitive alphabet. I have done some preliminary work and have needed alphabet but it also requires new go module to implement it - as logic is a bit different than current one.
Unless you have billions of files in some folders chances of problems are very tiny. But choice is yours.
IMHO I would say it is safe (with caveat that in theory not 100%). In computing we rely often on such solutions e.g. hashing rclone is using.
However when with hashing we gain something (speed) in trade off of 100% confidence and there is no known better algorithm in case of file names' base32768 encoding we know 100% safe solution exist and is equal in terms of complexity/speed as current one - if you are a go programmer and would like to help I am happy to share my work so far on it.