You're right, I was using 0-9A-Z, when it should have been 0-9A-F. I still need some coffee :)
So for 6,984 URLs per 32-bit hash, wouldn't that be evenly distributed since it's the result of a hashing function? Therefore we'd expect fairly close to 6,900 URLs per prefix? In what situation would you expect a 1-to-1 of 32-bit prefix to URL? Note: happy to be disproven, this is not my specialty at all.
We are working in bits, use bits instead, that will avoid theses kinds of mistakes.
> In what situation would you expect a 1-to-1 of 32-bit prefix to URL?
Oh yeah sorry I misunderstood it, yeah it's pretty unlikely that you would get 1 url for a prefix(but still possible). I would have to get out my old probability books to find that out but it's not worth it, the probability would be way too tiny.
I thought it was about certainty to be able to match it with the real URL. In theory it would takes only a few page hit to be certain of the domain and thus the URL (if there's no unknown string in the URL).
So for 6,984 URLs per 32-bit hash, wouldn't that be evenly distributed since it's the result of a hashing function? Therefore we'd expect fairly close to 6,900 URLs per prefix? In what situation would you expect a 1-to-1 of 32-bit prefix to URL? Note: happy to be disproven, this is not my specialty at all.