We have two systems that are kept synchronized through zfs send / zfs receive. Normally, the disk space usage on the two is identical, or at least extremely close.

Since installing 0.6.3 on the secondary server (was previously running 0.6.2), destroying the pool, and doing a complete re-sync with "zfs send -R", the space used is now quite different, about 10% higher as a rough guess. (Roughly 400GB out of 4TB).

I did notice that we had set "xattr=sa" on the source filesystems, and that attribute did not get replicated through the send/recv. But after starting over and setting xattr=sa on the destination filesystems, that made no difference.

Is there any reason that space usage on filesystems using 0.6.3 would be higher than 0.6.2?

As an aside, when blocks are sent and received, are they transferred in the native compressed format, or are they uncompressed, transferred, and then re-compressed on the destination? Just curious, thanks.