[Http-crcsync] General comments on crcsync document

[Patrick McManus]

it occurs to me that there is an obvious mitigation: any document with
100K hash values and just 1 hit is a really poor delta and indeed
probably isn't a delta at all. Any response that is _almost_ all
literals should probably just make itself into all literals (or just
plain 200-non-delta).. 

this would really change the math significantly for the better, and I
don't think you are giving up any use cases where the deltas will be
effective anyhow.

thoughts on that?


On Fri, 2009-07-17 at 13:32 -0400, Patrick McManus wrote:
> you've triggered an a-ha in me.
> 
> for some reason in my head I had the server calculating the same number
> of hashes as the client (e.g. 40).. but that's not true.. the server
> does a rolling calculation and, in the case of all misses calculates
> every possible window - which in the case of a 100K document is just shy
> of 100K different hash values.. which is a lot more than 40 ;)
> 
> do I have that right?
> 
> so 100k crcs each of which could falsely match 40 different client
> supplied crcs gives us 4 million lottery tickets into a 4.2 billion
> address space.. so that puts us on the same page at 1 in 1000 (modulo
> those bogus assumptions).
> 
> That's a whole lot of crcs ;)
> 
> 
> 
> 
> 
> _______________________________________________
> Http-crcsync mailing list
> Http-crcsync at lists.laptop.org
> http://lists.laptop.org/listinfo/http-crcsync