Re: new iSCSI draft - 02.txt

[Matt Wakeley <matt_wakeley@agilent.com>]

David Peterson wrote:
> 
> The point is (whether you use an mbuf/mblk/zbuf/socket buffer or whatever)
> some type of manipulation will be required. Bottom line is a periodic marker
> does not work well for a software based implementation. As I've indicated
> before
> I do not believe a framing mechanism is required at this point in time. We
> should start looking at what it will take to use SCTP (or something
> similar).

Perhaps it's not needed now by the first generation software implementations,
but it must be standardized now so that when it is needed by tomorrow's
hardware implementations, it will be there.  If you relegate the future to
SCTP, then that means that today's products (using TCP) will not be able to
interoperate with tomorrow's products (using SCTP).

As I have stated, the marker is negotiated, and only supplied if the receiver
requests it.

-Matt Wakeley
Agilent Technologies

> When the 10G pipes become available in the future we will then be
> ready/positioned
> and have a solution that should work for all.