Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!samsung!emory!utkcs2!moore From: moore@cs.utk.edu (Keith Moore) Newsgroups: comp.sys.ibm.pc.rt Subject: Re: RS/6000 survey - will post summary Keywords: Time to get some real-world opinions Message-ID: <1990May6.212646.20580@cs.utk.edu> Date: 6 May 90 21:26:46 GMT References: <11111@bsu-cs.bsu.edu> <50279@lanl.gov> <1990May4.222715.23648@cs.utk.edu> <10036@stiatl.UUCP> Organization: University of Tennessee CS Department Lines: 82 In article <10036@stiatl.UUCP> tom@stiatl.UUCP (Tom Wiencko) writes: >moore@betelgeuse.cs.utk.edu (Keith Moore) writes: > >>While release 9013 (+ n) might fix some bugs, it can do nothing about >>design flaws (they call them "enhancements"), and there are lots of them. > >>IBM has yet to understand that they don't understand. > >>AIX v3 is NOT UNIX. Ask anyone who has managed to use it for more >>than an hour. I've only spent a few hours trying to port various >>pieces of software to this box, but by now I start screaming after >>trying to use it for more than a few minutes. > >>Pray you never actually have to use this system. > >That is awfully funny ... I have ported a pretty good bit of code to the >RS/6000 and the only problems I have had are compiler and library bugs. >I have yet to run into any design flaws. None. Not one. Some kinds of bugs are evidence of poor coding or accidents. These are easily fixed. Others are evidence of design flaws. I know that something is seriously wrong, for example, when I do an "stty erase ^h", change shells, type "stty" again, and find that I have no erase character...from poking around it a appears that they have several parallel tty drivers to accomodate BSD/POSIX/SYSV/whatever, but have done so in such a way as to make a real mess. I'm sure it will get "fixed" eventually, but if the implementation were done right in the first place, these kinds of bugs would never happen. Or how about the fact that nearly all of the utilities now have verbose error messages, even when the old terse message was sufficient? After seeing a few *incorrect* verbose error messages, I started to form the opinion that the effort to change every message in the entire Unix software suite was misdirected. Of course they will fix the messages eventually...all someone has to do is submit a bug report for every incorrect or misleading message. Or how about that at sometimes one of our AIX machines lets you log in as "root" from a telnet connection without a password, even though one *has* been assigned with "passwd root"? (I'm sure IBM will fix this one soon, and I'm also sure that we somehow "caused" this problem because we tried to add a user to the system without understanding how the new security features of AIX break things that work in traditional UNIX.) What bothers me most is that all of the above, and nearly every bug I've seen, is the result of changes that have been made to software that *already works* as distributed by AT&T, UCB, or whoever...at least the copies on our source tapes work reasonably well. I've been using Unix systems for ten years...not as long as many, but enough to feel at home. When more than half of the commands I type on an AIX box don't do what I expect, I start to suspect there's more than just a few bugs that need fixing... >Perhaps you can be a little more specific about exactly what it is that >IBM does not understand. How about the Unix design philosophy? How about simplicity and elegance? How about "if it works, don't fix it"? >I do not have to use the system, but I really like it so far. It also >seems more "compatible" (that is, runs more things without messing around) >than most other flavors of UNIX I have run into. It's obvious that the AIX v3 developers tried very hard to make the system "compatible" from both the shell-command level and the C-callable-function-library level. But in doing so, they made the system far more complex than it needs to be, and introduced lots of failure modes that confuse experienced unix programmers. >Perhaps you just hate IBM and would like to take it out on this box? An easy shot, but it misses entirely. I have an RT (running 4.3) on my desk, even though I was offered a SparcStation. I like the RS/6000 hardware -- a far as I can tell, the box is well designed, and a lot of cycles/second/buck. But the operating system loses. The above is entirely my opinion. -- Keith Moore Internet: moore@cs.utk.edu University of Tenn. CS Dept. BITNET: moore@utkvx 107 Ayres Hall, UT Campus UT Decnet: utkcs::moore Knoxville Tennessee 37996-1301 Telephone: +1 615 974 0822