Xfs is still faster but ext4 is catching up on large file creation

That is a misunderstanding. All reasonably performing filesystems lose data if you pull the plug. The difference with XFS is that it buffers writes more aggressively than e.g. ext3, so the amount of lost data may be greater. Properly written apps fsync() any precious data anyway, so this is not an issue.

The notion that XFS is unreliable probably stems from the way such lost buffered data has historically been handled. Because XFS allocates the space ahead of actually writing the data, files with unwritten data at a reset could have visible zero-filled blocks after recovery whereas ext3 would simply truncate them. In both cases the data was lost.

Since a while back, unwritten extents like these are detected and trimmed back, so now XFS behaves more like most other filesystems. There are still more buffered writes floating about, of course, so the same precautions are still required.

What would that be ?

I am really interested to know. Right now I am running EXT4 on an SSD in my notebook. It performs well. Sometimes, after a suspend , I get a load of EXT4 error messages which are stalling the whole notebook. Hard resseting it, makes the error disappear. Up to now, no data loss.But anyway, this makes me a nervous and I am not sure if it is due to Kernel, SSD, Ext4 or the combination of all/any of them. I guess XFS is way out of my use point, but anyway, an option is an option.

Well quite - ideas that others are struggling to match in 2011.