[Http-crcsync] General comments on crcsync document

[Patrick McManus]

On Sat, 2009-07-18 at 10:51 +0200, Toby Collett wrote:
> 

> 
> The problem here is that a non buffering implementation of a chunked
> transfer doesnt know it has only got one hash match before it starts
> sending the data...
> 

what I was thinking was more along the lines of "I haven't seen a hash
match even after calculating 4(?) full windows worth of data - I'll stop
looking because even if this is a legitimate delta it is going to be a
badly performing one".. 

that bounds the number of hashes so should really help with the false
positive problem and doesn't pose a problem for streaming the response..