I’ve had a problem with my Apache setup where it couldn’t serve images with unicode characters in file names:
http://www.boku.ru/img/проверка.jpg (really missing right now; just an example)
SCP client (WinSCP) clearly showed files were present, other files from the same folder could be accessed but those with cyrillic characters were returning 404.
Using Wireshark, I determined my browser sends urlencoded text (
B0%D0…) instead of bare unicode, but that was, apparently, like it should be.
Studying Apache logs, I’ve found it complain about missing files, re-encoded in
\x notation (
\xb0\xd0…). For a while I thought this suggested a problem, but in the end it was unrelated and probably correct behavior.
Turns out, WinSCP does not use UTF-8 by default. Instead it sends file names to server in the local PC encoding, Win-1251 in my case. Looking through WinSCP everything seemed alright because it reads filenames back in the same encoding. But from the server point of view (as seen through SSH for example) those were not Russian characters but some Krakozyabry.
In other words, there really was no
To recongifure WinSCP, open Site manager, press Edit, Advanced, “UTF-8 encoding for file names: On”, then reconnect. You’ll have to rename all affected files.