I don't see the orders of magnitude. The Figure 13 shows an average
increase of 200-300x and my estimate was 87x for F=10 or 870x for F=1.
It's not clear to me how many servers did you assume in your
calculation the client was using normally and by Khronos, but there
seems to be an agreement.
Got it. I thought you meant 8700x.
With the 10x interval suggested in the Khronos draft the increase for
the slowest servers would be still 20-30x by your estimates. I don't
think that is acceptable.
I think that we are in agreement about what the security-overhead tradeoff looks like, and that any balance
between the two can be struck depending on the specific choice of parameters. For instance F=100 and only
considering servers with netspeed >=10 would still leave us with >70% of the servers (Figure 12) while leading to
a ±100x reduction in the load per server with netspeed = 1. I propose adding to the draft an explanation how the choice of value assignment
to the parameters affects the load per server.
As a side note, I don't think that the acceptable/unacceptable borderline should merely be defined by the multiplicative
factor of increase, but also by absolute numbers (average #bytes per second). That said, once again, any target
absolute upper bound on load per server can still be reached through the proper choice of parameters. We will clarify that in the text.
-- last-call mailing list last-call@xxxxxxxx https://www.ietf.org/mailman/listinfo/last-call
