Speed seems OK for me. However since about 2.1.22 I have been occasionally getting the following:
bsd_decomp0: bad sequence # 151, expected 148
bsd_decomp0: bad sequence # 67, expected 65
bsd_decomp0: bad sequence # 57, expected 55
bsd_decomp0: bad sequence # 183, expected 181
bsd_decomp0: bad sequence # 1010, expected 1008
bsd_decomp0: bad sequence # 83, expected 81
bsd_decomp0: bad sequence # 103, expected 101
bsd_decomp0: bad sequence # 309, expected 307
bsd_decomp0: bad sequence # 261, expected 259
bsd_decomp0: bad sequence # 69, expected 67
TCPv4 bad checksum from 04101dcb:0014 to 80101dcb:07e3, len=1480/1480/1500
TCPv4 bad checksum from 04101dcb:0014 to 80101dcb:07e3, len=1480/1480/1500
bsd_decomp0: bad sequence # 83, expected 81
bsd_decomp0: bad sequence # 349, expected 346
bsd_decomp0: bad sequence # 76, expected 74
bsd_decomp0: bad sequence # 319, expected 317
bsd_decomp0: bad sequence # 264, expected 262
TCPv4 bad checksum from 0eec0ccb:1f90 to 81570cca:0448, len=532/532/552
TCPv4 bad checksum from 0eec0ccb:1f90 to 81570cca:0491, len=532/532/552
TCPv4 bad checksum from 04101dcb:0014 to 80101dcb:0497, len=1480/1480/1500
When the bsd_decomp errors start occuring the TCP session in question becomes stuck. I make it work again by using "strobe" on the remote host which clears out all compression buffers...
This generally occurs on big transfers, on versions 2.1.25 and 2.1.26 FTP mirroring was almost impossible because of this.
If any networking person has some tests they'd like me to perform then just say so.
Russell Coker