You are not logged in. Login Now
 0-3   4-28   29        
 
Author Message
cross
Grex: It's time to switch operating systems. Mark Unseen   Nov 29 20:48 UTC 2010

So after fighting the OpenBSD package system for the last week or so, 
I'm about to give up.  The stuff just doesn't work.  OpenBSD sucks.

Background: When Grex was founded, and for the first few years of its 
existence, it ran on Sun hardware using Sun's operating system, 
SunOS.  At the time, SunOS was a derivitive of the Berkeley 
distribution of Unix (Sun later switched to AT&T's System V Release 4, 
which is more commonly known as "Solaris" or "Solaris 2", but formally 
known as SunOS version 5 [SunOS 1 through SunOS 4.x were the BSD 
derived versions; actually, the earliest might have been 7th Edition 
derived, but that's neither here nor there and really getting into the 
weeds on something that's only tagentially related].  Confusingly, as 
a marketing tactic, they later retroactively named the older, BSD-
derived SunOS Solaris 1).  The hardware originally used Motorola's 
68000-derived family of processors, but eventually switched to the 
RISC SPARC processor.

Around 2002 or 2003, Grex's hardware was starting to break down, the 
BSD-derived SunOS had been EOL'd by Sun, and the price/performance 
point for Sun hardware was really being eclipsed by x86 hardware.  
Further, there were several free, very stable and mature operating 
systems available for x86 hardware.  It became clear that it was time 
to at least upgrade; after a lively round of debate, it was decided to 
use an x86-compatible AMD processor, new SCSI disks, etc.  Then the 
debate around operating systems started in earnest, with the field 
eventually narrowed to two reasonable candidates: OpenBSD and 
FreeBSD.  Of these, FreeBSD has a much larger userbase, much better 
third party software support, and is generally more advanced.  OpenBSD 
focuses more on software security and correctness, with many 
developers actively auditing the source base looking for (and often 
finding and fixing) security problems before they become real 
problems.  Of note, and relevant to this discussion, is that FreeBSD 
does the same thing; also, something that's come up recently but that 
was not, and is not now a consideration, is that M-Net runs FreeBSD.

Marcus and Steve both argued passionately for OpenBSD on security 
grounds, I argued for FreeBSD on technical grounds, everyone else was 
ambivalent.  Ultimately, the decision was made to go with OpenBSD.  
Hardware was purchased, and then ... nothing happened for a year.  The 
hardware sat in somebody's house with no one really looking at it.  
Finally, Jan Wolter took the initiative to get Grex up and running on 
the new system; he developed a less than favorable impression of 
OpenBSD along the way (see some of the old archives in the garage 
conference).  Joe Gelinas and I did some additional work to get things 
going, someone (probably Jan) moved the data over from the Sun and put 
the new machine into production in December of 2004; it's been running 
ever since, modulo some new hardware to replace failed components 
(particularly hard discs) and upgrades of the operating system.

Since sometime in 2007, I've been what you might call the primary 
staff guy, in the sense that I've done most of the upgrades and a lot 
of work on the software, the web site, etc.  In that time, I've formed 
the opinion that OpenBSD was a mistake, and that we really, truly 
would be better off with FreeBSD.  I want to bring this back up for 
discussion now.

Some basic technical points:

a) Grex's current hardware is getting long in the tooth and needs to 
be replaced at some point.  Modern computers are based around 
multicore processors; OpenBSD doesn't support these particularly well; 
certainly not as well as FreeBSD.

b) I do not believe that OpenBSD is any more secure than FreeBSD.  In 
fact, we've seen instances of OpenBSD security holes *on Grex* that 
didn't appear in FreeBSD.  Many of the flaws that Chad and Mickeyd 
used to crash Grex continuously don't appear in FreeBSD: despite 
trying to do the same things on M-Net, that system remained stable.

c) Many virtual hosting providers support FreeBSD, but relatively few 
support OpenBSD.  Grex has probably got one more round on its own 
hardware before going virtual and living in the cloud.  It would 
behoove us to transition to a software stack that's going to make that 
transition as painless as possible.

d) FreeBSD has a LOT more third party software support.  Things that 
could be a big draw for potential users exist in the FreeBSD ports 
tree, but not OpenBSD; sure, those things *may* eventually get ported 
to OpenBSD, but why wait?

I can go into more depth, but I think it's time to get serious about 
switching.
29 responses total.
kentn
response 1 of 29: Mark Unseen   Nov 29 22:22 UTC 2010

Thanks for the overview, Dan.  It seems like a good idea to me to use
whatever OS that staff feel is best, that supports the applications we
currently run, applications we want to run, and an OS staff would like
to continue working on.  We definitely need something that will make
life easier for staff as well as something that makes it easier for Grex
to provide the services our users want.  And being compatible with 
the offerings from hosting providers sounds like a plus.  I'm curious
what other people think.
jep
response 2 of 29: Mark Unseen   Nov 29 23:08 UTC 2010

I think most of us are going to make, or go along with, any decisions on
the basis of non-technical information.  I don't know enough about the
differences between OpenBSD and FreeBSD to make any difference.  Neither
do most of the users, or most of the Board.

I would be willing to bet there are two or three people involved in
running Grex who should decide such things, and the rest of us will go
along.  I'm interested in details like whether the slime who hack Grex
and bring it down can be stopped, not how they can be stopped.  If there
are cool capabilities to be gained, I would be interested in what those
are.  Can Grex be opened up again so people can log in and use it right
away, the way I did when it first opened to the public?  I want it open
again, and I want it as soon as possible.

How long is it going to take to transition to FreeBSD, once the decision
has been made to make that move?  I mean in weeks on my calendar, not
hours of staff time.  If you can do it in a week, great. Start this
Sunday and we'll see you in a week.  If not... then how long?  Are you
going to do it yourself, Dan?  If not, do you have the help you need?

How long until Backtalk will be available?  It's what I use here, so
it's what I care about.

Also, is there any cost?  How much?  Are there users who will pay for
it?  As I said elsewhere, I can chip in some.

Tell me things like that.
nharmon
response 3 of 29: Mark Unseen   Nov 30 02:23 UTC 2010

Users probably would not see much a difference between FreeBSD and
OpenBSD, so I think whatever is easier for staff is the way we should go.

I might also mention that Dan really does seem to be our main staffer
right now, if not the only active staffer. I think it would be a good
idea for the board to give him some authority to take charge of
technical operations and make decisions like this. 
 0-3   4-28   29        
Response Not Possible: You are Not Logged In
 

- Backtalk version 1.3.30 - Copyright 1996-2006, Jan Wolter and Steve Weiss