- From: Chuck Shotton <cshotton@oac.hsc.uth.tmc.edu>
- Date: Wed, 7 Dec 1994 09:49:38 -0600
- To: hallam@alws.cern.ch, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
At 8:17 AM 12/7/94, hallam@alws.cern.ch wrote: >To sum up Marcs argument: > >1) The performance hit is not too great This is a fallacy. Simply put, it comes down to whether or not every single character in a text transfer must be examined. There are N comparisons that must be performed to place text in a cannonical form (one for each byte in the file) versus ZERO comparisons if clients and servers simply tolerate multiple line ends. How anyone can say this doesn't impose a noticable impact on a server is beyond me. To carry it further, simply write a C program that compares execution times for block oriented reads from disk with no conversions versus character by character I/O with CRLF conversions. The performance difference is worse by almost 200% for the latter. Try it yourself. >2) If there is no reason to do it and no reason not to then follow the spec. > >I do not want cannonicalisation under any circumstances. I have had my fill >of systems that "canonicalise" trying to be "clever". Such systems break >much much more than they mend. I agree 100%. My whole point in participating in this thread has been to advocate adding some verbage to the standard that documents the current practice of tolerating mixed line ends as the prefered method. That's it. Somehow this "cannonicalization" side show crept in and stirred up a hornet's nest. --_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_\_-_-_-_-_-_-_-_-_-_-_-_-_-_-_- Chuck Shotton \ Assistant Director, Academic Computing \ "Shut up and eat your U. of Texas Health Science Center Houston \ vegetables!!!" cshotton@oac.hsc.uth.tmc.edu (713) 794-5650 \ _-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-\-_-_-_-_-_-_-_-_-_-_-_-_-
Received on Wednesday, 7 December 1994 07:51:54 UTC