Loading free software on Suns [LWN.net]

Loading free software on Suns

Posted Nov 18, 2005 0:47 UTC (Fri) by corbet (editor, #1)
In reply to: Sun Announces Support for Postgres Database on Solaris 10 by mikec
Parent article: Sun Announces Support for Postgres Database on Solaris 10

> Perhaps that was obvious to everyone else long before I noticed it? :-)

It was Larry Augustin, as I recall, who commented on what was a common experience fifteen years ago: we would buy palletloads of Sun systems, and immediately replace everything we could with free software. You youngsters, who have never been through the experience of trying to build early X11 releases with early gcc releases, just don't know how good you've got it.

It was painful sometimes, but well worth it, even back then. A system without free software just feels dead.

(Log in to post comments)

Loading free software on Suns

Posted Nov 18, 2005 2:41 UTC (Fri) by mikec (guest, #30884) [Link]

Well, I did not manage to avoid that pain - I just found it elsewhere trying to compile X11 for 64bit alpha and dealing with (void *) typing issues in drivers...

I just did not try to do anything with a Sun Box until more recently (well almost 5 years ago now - wow has it been 5 years?)

I always assumed all the pain I suffered on Alpha,Linux and the like at home though the early and mid 90's - relative to the stability of Sun platform for development at work - was that Sun really had an edge (at least in the late 90's) - it was only when I realized how hard my Sun sysadmins were working to hide this from me that I understood what was really going on...

Of course the sad thing is that even with all that toil and "instability", I still managed to have 'nix boxes at home running for months on end while the windows boxes needed rebooting daily...

Loading free software on Suns

Posted Nov 18, 2005 2:49 UTC (Fri) by mikec (guest, #30884) [Link]

Oh and building 2.1.x for Alpha kernels weekly, recompiling milo - good times :-)....

It's a good thing that processor was so much faster for floating point computation, or between the reboots, I would not have gotten anything done...

Yeah, yeah, we know:

Posted Nov 18, 2005 15:25 UTC (Fri) by Baylink (subscriber, #755) [Link]

10 miles.

In the snow.

Uphill.

Both ways.

;-)

I go back to 1982, when a denizen of my local JC was trying to *write* a MUA from scratch, for our SWTPC 16-user 6809 box.

We got a trio of AT&T 3b2-300's (remember those), and he gave up.

Loading free software on Suns

Posted Nov 18, 2005 19:58 UTC (Fri) by tjc (subscriber, #137) [Link]

It was Larry Augustin, as I recall, who commented on what was a common experience fifteen years ago: we would buy palletloads of Sun systems, and immediately replace everything we could with free software.

Yeah, I remember that. Sun shipped a buggy version of yacc for years. It was broken, and I assume they knew it was broken, but they shipped it anyway. Most people who needed a parser generator replaced it with GNU bison.

I've forgotten how bad bare Solaris is

Posted Nov 19, 2005 3:09 UTC (Sat) by zblaxell (subscriber, #26385) [Link]

The memories are fading rapidly. They're just too painful...

On Solaris I could never remember whether it was the stuff in /usr/bin, /usr/ccs/bin or /usr/ucb/bin or /usr/sys5/bin that should never, *ever* be used, because it was buggy in some subtle and pervasive way--instead, the stuff that immediately and obviously segfaulted or aborted should be used, because at least that stuff produced *correct* output when it worked. I think it might have actually been dependent on whether it was SunOS 4 (aka Solaris 1) or Solaris 2 (aka SunOS 5) or...

I do remember that the machines were unusable until I had replaced almost everything in /usr/bin with GNU tools--and then it was usable, but don't expect *everything* to work. I also remember that the system would explode if I tried to actually replace the stuff in /usr/bin with GNU tools. I spent much of my time making sure that the environment had $PATH a) defined in all cases (X login, ssh, telnet, rsh, rexec, ...), and b) never contained /usr/bin. It would have been much easier if I could just nuke /usr/bin or replace it with *maintained* software, without breaking all the applications on the machine.

Still...from what little I do remember, out-of-box HP/UX is much, much worse. It's like a time capsule from 1985. Segfaulting, POSIX-violating, utterly unusable, broken libc functions--code that could have no legitimate users since it could do nothing other than "kill(getpid(),SIGSEGV)"--were not fixed after several years and major HP/UX releases, despite the existence of HP patches to several of those major releases to resolve these problems. HP seemed to distribute new features bundled on the OS install disc, but every bug found in the last two decades was distributed as a patch. It was like they were afraid of changing their own binaries or something.

I've forgotten how bad bare Solaris is

Posted Nov 19, 2005 18:22 UTC (Sat) by mikec (guest, #30884) [Link]

HPux is so bad that each time I had to deal with it, six months later I had forgotten (thankfully) that I had...

That one is so bad that:

a. you wonder if anyone really used it at all?
b. you have to read the man pages from other OSess to figure out what the !@#$ is going on...