Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!fletcher From: randyc@hpfclj.hp.com (Randy Campbell) Newsgroups: comp.std.unix Subject: Re: need POSIX cksum tables Message-ID: <11189@cs.utexas.edu> Date: 14 Aug 90 23:18:04 GMT Sender: fletcher@cs.utexas.edu Reply-To: std-unix@uunet.uu.net Lines: 38 Approved: fletcher@cs.utexas.edu (Guest Moderator, Fletcher Mattox) From: Randy Campbell >From: andrew@alice.uucp (Andrew Hume) > > i would hate to cause trouble but can anyone justify >why cksum uses anything other than CRC-32 for its polynomial? >(i am not 1000% sure it isn't but am fairly sure.) >it really frosts my shorts when CCITT and other folks >bust their chops developing and testing 32 bit CRCs, find a really good >one and then posix picks some hokey thing whose only apparent recommendation >is that pathalias uses it (not strictly the highest form of praise). > > >Volume-Number: Volume 21, Number 23 >---------- Well, I had some part in developing the current (Draft 10) spec of cksum. There was certainly no intention to "frost your shorts" (is this a reference to some kind of stone-washed jeans?). The Draft 9 spec for cksum (the one that was derived from pathalias) had a problem in that it would too often yield identical results for files that were different only near the beginning. It may be that CRC-32 would be an adequate or even superior polynomial to the one we used. However, the polynomial selected is the one used by Ethernet and is specified in a networking standard (ISO 8802-3), so we thought it would be appropriate to use. In fact, finding a known, standard polynomial to use was one of our criteria. And it gave good empirical results on some filesets that had exposed weaknesses in other implementations. Randy Campbell Volume-Number: Volume 21, Number 31