Resuming Upload with APPE command creates corrupt files

Hi,

uploading, aborting and resuming uploads creates corrupted files on some servers. WinSCP uses the APPE command to resume, FileZilla uses REST+STOR.

Some time ago FileZilla also creates corrupted uploads, but this is fixed now.

Use REST+STOR to resume upload instead of APPE if server advertises REST STREAM support

Is there a chance that WinSCP supports the REST+STOR command if advertised by the server too?

Thanks for your post. This request has been added to tracker.

Can you maybe raise the priority a bit for this feature? E.g. if I upload a 1,5GB file with 60KB/s and after 5h there was a short connection interrupt, it sucks to restart upload again, during other clients are able to resume upload. Especially raising priority would be nice, because it seems that resuming doesn't work with ProFTPD with enabled APPE and REST support.

icke wrote:

Can you maybe raise the priority a bit for this feature? E.g. if I upload a 1,5GB file with 60KB/s and after 5h there was a short connection interrupt, it sucks to restart upload again, during other clients are able to resume upload. Especially raising priority would be nice, because it seems that resuming doesn't work with ProFTPD with enabled APPE and REST support.

Ok, I've raised priority.

Use REST+STOR to resume upload instead of APPE if server advertises REST STREAM support

I was looking at this change in FileZilla and while I admit that using REST instead of APPE is semantically more sound, I do not think that it can have any actual effect on the transfer technically.

With APPE: Client (FileZilla/WinSCP) checks remote file size, tells FTP server to start appending to the remote file, and feeds data from local file starting with position based on remote file size.

With REST: Client (FileZilla) checks remote file size, tells FTP server to restart transfer, starting at a position equal to the file size (=end of the file), and feeds data from local file starting with position based on remote file size.

Technically, both approaches are identical. Or does anyone see a difference?