I've been trying to come up with a good pseudonym for the organization that I am the technical director for so that I can talk about it here without confusing what I am writing with whatever I am doing in my real job for Sun. How about Unnamed Social Network Site... or USNS for short? I suppose it really does not matter and that is easy enough to type.
 
A friend of mine donated a spare 1U server to USNS recently. We are currently leasing a server with our co-lo and it is not meeting our needs anymore. The donated server is not "new", but it was free and it should meet our current workloads. Like most non-profits, we operate on a shoestring budget and not having to pay to lease a server every month, especially one that cannot handle peak usage, is a big improvement for us.
 
Step one for the migration from the old server to the new server (and from Ubuntu to OpenSolaris) is getting the new server ready to be taken to the co-lo. Since the co-lo is on the other side of the country and we have no remote console set up, I need the OS installed before the machine is delivered to them. One of the members of my technical team, who lives in the same area of the country as the co-lo, will be doing this for me and then delivering the machine. It will be a great test of all the changes that have been made to the download and install process for OpenSolaris. I think the process is really easy, but I've been working with Solaris for a long time. Will someone with no Solaris/OpenSolaris experience see it the same way? I am about to find out. If all goes well, we'll have the new server online by the end of the month.

 
I had a really good/funny story about Tom's new job here. He started a new job, still at the University but now working for the group that supports the whole campus. He was hired to be their Solaris expert. The CIO of the campus used to be Tom's boss in his old job so he knew what Tom could do. I am thankful he took the job because I think if he had not they would have eliminated Solaris completely. The problem with where he works is they don't know Solaris and so they do not know how to do things correctly. The result is they have made a lot of mistakes which make the systems difficult to admin and makes them run less than optimal.  
 
He made me take the funny story down but that is ok because it was the lesson the story teaches that made me post it. Not knowing they are doing things wrong, they are left just thinking Solaris is not as good as Linux. So, instead you get our conversation about why I posted it:  
 
Kristin: you make great stories, http://blogs.sun.com/kamundse
Tom: don't do that
Kristin: no one from (your work) reads my blog
Kristin: ok ok, i think it is funny, and useful
Tom: it is, but I have to work here
Kristin: i dont just put it on because it is funny, it is important
Kristin: these guys could ditch Solaris because they think it sucks
Kristin: because they can't use it right
Kristin: so why is that? why are they doing it all wrong?
Tom: they are used to Linux, they want to force solaris to work like linux and they don't want to learn the advanced solaris tricks
Kristin: yes
Tom: hence the /etc/hosts file that solaris doesn't deal with correctly
Tom: not using the solaris patch process, but rather making it look as much like linux as they can
Tom: add to that disabling auto_home
Kristin: i think Sun people need to know how the system are being used
Kristin: they need to know how people expect things to work and then we either need to provide linux-like interfaces or really get the message out about how to do things right
Kristin: ok... can i put this on my blog?
Tom: ok

 
 
I've been told that for Linux, efficiency trumps all other attributes. For OpenBSD, security seems like the number one attribute. Something like Exokernel would put flexibility at the top. For Solaris, robustness and compatibility are good candicates for being in the top three. Other operating systems have attributes they explicity don't care about. Plan 9 creators decided against compatibility (with UNIX) in favor of doing everything over the "right" way.  
 
So, what do you think the top five (increased from three) attributes for Solaris, Linux (you can even break it down by distribution), OS X, the various BSDs, and any other OS you are interested in? This is a list of the top five design goals of the OS, not necessarily what attributes it actually displays.  
 
Here's a list (not complete) of some attributes to consider:

• robustness
• maintainability
• security
• interoperability/compatibility/standardization
• useability
• efficiency/performance/speed
• portability
• flexibility
• scalability
• correctness

 
 
Here's my first pass, I'll update this based on any comments.  
 

Solaris	robustness, compatibility, scalability, security
Linux	performance (changed from efficiency), portability
FreeBSD	compatibility, robustness, maintainability
OpenBSD	security, portability, standardization, correctness
NetBSD	portability, interoperability, security
OS X	useability, robustness
Generic μkernel	flexibility, useability
Windows	profits, lock-in, global domination (from comments)

 
 
If an OS is not here, its because I was not able to come up with anything I felt comfortable with. Oh, and I am not only talking unix-like OSes. I did not leave Windows out intentionally, I just don't know enough about it to know what its design goals are.  
 
For anyone who may be following it, I will get back to the 64-bit processor series in the next post.