[jappu]

For as much promise there was for these new devices, there has also been some
disappointment along the way. Such as it is for any new technology, but here's my
opinion of what went wrong, and what could be done to correct it.

As usual, the primary problem is one of legacy requirements. The SATA interface
isn't really an ideal interface for such a device due to the relatively slow transfer speed
and the constraint it imposes on the drive that the intelligence must reside on the
drive itself. The LBA addressing scheme used by SATA further isolates the drive
from knowing any details about the file system, and thus being unable to
efficiently manage data storage given the constraints of NAND flash memory,
such as page sizes, blocks size, wear leveling, and so on.

The next legacy problem is the file system. Rather than create an enirely new
file system designed to work ioptimally with SSD, we are forced to use the same
file systems that were meant for hard drives.