From mouse@collatz.mcrcim.mcgill.edu Sun Dec 18 16:02:19 1994 Return-Path: Received: from Collatz.McRCIM.McGill.EDU by life.ai.mit.edu (4.1/AI-4.10) for /com/archive/cube-lovers id AA14014; Sun, 18 Dec 94 16:02:19 EST Received: (root@localhost) by 13839 on Collatz.McRCIM.McGill.EDU (8.6.8.1 Mouse 1.0) id PAA13839 for cube-lovers@ai.mit.edu; Sun, 18 Dec 1994 15:56:10 -0500 Date: Sun, 18 Dec 1994 15:56:10 -0500 From: der Mouse Message-Id: <199412182056.PAA13839@Collatz.McRCIM.McGill.EDU> To: cube-lovers@ai.mit.edu Subject: Re: How Big is Big? > [Physicists] are planning soon to start sending petabytes (10^15) > over the Internet. 10^15 is getting interesting close to the size of > Rubik's cube (never mind that I thought that the proper term for > 10^15 bytes was terabytes.) I thought it was kilo 10^3 mega 10^6 giga 10^9 tera 10^12 peta 10^15 exa 10^18 except, of course, that as applied to quantities that tend to come in powers of two, like bytes, they normally refer to 2^10, 2^20, 2^30, 2^40, 2^50, and 2^60 instead. (This is a common problem when buying disks: manufacturers like to quote capacities in terms of powers of ten, because it makes their disks seem larger than they really are. A "2.1G" disk, for example, typically has a capacity of about 2.1e9 bytes...which is really only about 1.956Gb. The error can be roughly estimated as 2.5% per power of 10^3: 2.5% for K, 5% for M, 7.5% for G, etc. Semiconductor memory manufacturers generally get this right, probably because doing other than powers of two would be hard for them.) It also means that a certain well-known manufacturer of data drives for 8mm videotape is being extremely arrogant with their choice of name. :-) As for the 10^18 bytes of storage estimated (probably only about half that, if we consider that we really need only 5*.9e18 bits, less if we resort to some of the clever coding tricks recently mentioned)...that's about a gig of storage each across a million machines. The net's not quite to the point where it can be done distributed. Yet. :-) Incidentally, someone mentioned that you need only store two bits, or even only one if you don't use H turns, per position, because you don't need to know more than how to get closer to Start...and then said that it would be nice to have the full depth available nevertheless. If you have this enormous database of .9e18 positions available in the compact form, all that's needed to get the full depth for a position is to take the short walk through the tree back to Start. Also note that the Cube database storage size requires the highest prefix we have. Time to get SI to think up some more, I guess :-) der Mouse mouse@collatz.mcrcim.mcgill.edu