I drafted this yesterday, but now that I've revised it, I feel it needs a thread of its own.
I agree that commercial UNIXes are dying and that Linux will be their successor. My boss and I have been sorely tempted to nuke every single HP-UX machine CAEDM has and replace them with Debian/PA-RISC -- we've already got one test machine running it. What do I love about Linux and BSD that I dislike about HP-UX, IRIX, AIX, Tru64 and Solaris? Consistency. I've come to take for granted the consistency of the GNU utils and other programs. I hate running "tar zxvf $filename.tar.gz" and having tar remind me that it doesn't understand the "z" switch, and that $filename.tar.gz isn't a valid tar archive anyway. I hate ssh'ing to a machine and never knowing if hitting <backspace> on my keyboard is actually going to remove the character before the cursor. I hate trying to get an app written for gcc to compile with headers that are parseable only with the system's proprietary compiler (like getting blender to compile against HP-UX's OpenGL libraries so I can take advantage of the $1500 graphics card in my machine). I hate forgetting that the tape I'm trying to read was written by an IRIX machine, and that IRIX's tar uses a blocksize different than everyone else's, so I'm going to have to manually force the block size to 1k and NULL-pad the output so that tar reads the end of the tape. I hate using find and grep to search through the entire /var tree to find the output of the system logs, only to discover that they're stored in /usr/local/sys/adm/log, and I didn't know this because the syslog.conf file is in /adm/priv/conf/etc . I hate discovering that there's nothing "Common" about Solaris's CDE and that I'm going to have to visually parse their config files just to change xterm's default font (because for some reason, it won't read the default font resource setting out of my .Xdefaults file). I hate having to refer to a manpage just to see what the programmers decided to use as their command-line switches for some program in this particular release of Tru64. I hate discovering that I need a driver for some arcane hardware, and the only driver exists for an OS two releases ago, which is not binary-compatible with the current version -- and of course there isn't any source code. Yes, there are differences between Linux distros and Linux and BSD. But they aren't nearly as bad as the differences between commercial UNIXes, and sometimes between releases of a single OS. And therein is the commercial UNIX's weakness, and will be the chief contributor to their downfall. Software developers will no longer stand to have to write different code for different platforms when porting should be as easy as running "./configure && make all". Sysadmins won't stand to have to re-learn a new OS just to change the static IP address of their ethernet card. Commercial UNIXes don't *have* to die just to achieve this consistency, but they probably will anyway. Unless their developers contribute a lot back to the Linux code base, we stand to lose a lot. We'd lose Solaris's unmatched threading performance -- which would be a real boon to all the Apache + MySQL setups everywhere. We'd lose IRIX's CPU-scaling and memory architecture, not to mention its graphics subsystem which makes XFree86's DRI look pretty pathetic. We'd lose AIX's solidity and Tru64's VMS-derived clustering and partitioning features. Fortunately, IBM and SGI are making efforts to prevent this loss from occuring, but there is always the tradeoff between performance and portability. I don't see the death of commercial UNIXes as been necessary -- note that Debian now runs on Linux, FreeBSD, and NetBSD based-systems, so theoretically could run on Solaris, IRIX, or AIX based-systems -- but I do see it as being very likely. -- Soren Harward [EMAIL PROTECTED] ____________________ BYU Unix Users Group http://uug.byu.edu/ ___________________________________________________________________ List Info: http://uug.byu.edu/cgi-bin/mailman/listinfo/uug-list
