RE: [LinkSec] RE: Reminder: linksec call on Tuesday at 2:00pm ET
Mick,
A quicky, since I have the Linksec call in 8 minutes..
> I really enjoy your emails, they bring some life to a subject
> that I feared would be deadly dull.
You're most welcome. For certain, dullness will come later.
> Key
> retrieval performance may be only an insignificant
> contribution to overall transit delay.
>
My intended point was a little more subtle. The effect of key retrieval
time on throughput should generally be zero. However the difference
between no buffering in the front end of the MAC and a tiny bit of
buffering at the front end of the MAC (E.G. to allow time for key
retrieval) is a huge increment in MAC complexity. You now have
'pipelined packet state' in the MAC and all sort of corner cases, timing
overlaps and other nasties can crop up. Making the data path from the
front of the MAC to the intermediate storage (E.G. the host interface
buffer) combinatorial is a great simplifying goal in MAC design and it
would be a good thing if the standard facilitated this.
You will notice that I rant about implementations issues quite a bit. I
guess I'm not sufficiently theoretical.
Regards,
DJ