Re: [Dav-linuxfs] regression with versions after 1.2.1

[Werner B. <wer...@on...>]

The problem is caused by an error in the fuse kernel module. It 
miscalculates the allowed size of data in writes, which results in to 
big messages resulting in an IO-Error. The davfs2-version in CVS now 
takes care of this error by telling fuse a maximum size for write, that 
is smaller than the actual buffer size, and the error no longer happens.

The problem occurs with all 1.x.x-versions of davfs2. It only did not 
show up, as long as davfs2 used coda by default. With version 1.2.2 
davfs2 switched to use fuse by default.

To use fuse or coda:
The main difference (for davfs2) is: in read and write operations coda 
acts directly on the cached file, without involving davfs2. If you 
mainly upload big files, coda should show slightly better performance.
When the contents of directories is changing frequently, coda seems to 
sometimes reuse old, stale information, causing "file does not 
exist"-errors. fuse does not show this problem.


Cheers
Werner