Re: iSCSI has already been done before, sort of...

[Matthew Jacob <mjacob@feral.com>]

[ emerging from lurking - I'll probably regret it .. ]

You know, I actually was seeing this myself also. In fact, we implemented much
the same thing at NASA/Ames with IPI-3 over TCP/IP (at first) and then planned
for a frame mode HIPPI implementation.

While it's true that IPI-3's symmetric request/response model might be better
suited to a network transport that iSCSI is becoming, one of the main things
that iSCSI should be able to do is to (re)capture existing platform
implementations- i.e., use all of this (supposed) investment in SCSI target
device drivers. An IPI-3 implementation 

The trouble with IPI-3 is that the development of command sets for it stopped
around 1995 or so, so it has not kept up with a lot of the nifty features of
SCSI since then (SES && WWNs, to be precise).

-matt