#social 2019-03-14

2019-03-14 UTC
vitalyster left the channel
#
dzho cwebber2: wouldn't a bloom filter be used in a multipass approach? Narrow the set of possible positives, but do a more exacting search only on that set?
#
dzho doesn't really know these things at all
emacsen, xmpp-social, vitalyster and timbl joined the channel
#
cwebber2 dzho: a bloom filter is just a sizable bytevector, where hashes are made to the size of that large bytevector, and then are bitwise OR'ed into a union where each 1 is where *any* member of the set had a 1 in its hash and a 0 is where no members had a 1
#
cwebber2 dzho: so to narrow the set of possibilities
#
cwebber2 you still have to iterate over *every* possibility to check if they might be in there first
#
cwebber2 which is why I'm saying it's O(n) all possible iterms
#
cwebber2 so then:
#
cwebber2 - you could do a multipass approach, but then you'd still to keep track of all the delieverable members, in which case you're sending and tracking that data anyway... what are you gaining?
#
cwebber2 - if you don't do the multipass, you might misdeliver to some people
#
cwebber2 I don't see any wins out of this
#
cwebber2 but I think Gargron was playing with the idea and not committed to it
#
cwebber2 I thought it was interesting so I was just thinking it through :)
#
dzho fair enough!
cwebber2, jglauche, naturzukunft, dmitriz, DenSchub and timbl joined the channel; vitalyster left the channel