For serving as a music server at home, by far the easier way is to use iTunes and iTunes sharing. That allows you to play music from any Mac/PC/iPhone/iPad/iPod Touch/Apple TV on the same network. You can also set up home sharing which allows you to additionally copy the music around different computers. iTunes also works for movie sharing. The only downside of this is that it only works if the music/movies are in a compatible format, which explicitly excludes Windows Media, Ogg Vorbig/THeora and DivX.

For pictures, just use iPhoto. You can access local libraries the same way as with iTunes from any other Mac.

If you run iPHoto and iTunes on the server, you can use either iTunes/iPhoto on all clients or you can use Frontrow for browsing everything from within one application.

I just bought a mac mini for my server, but haven’t bought the server software yet. I am just looking for a cheap solution and I apologize now I am completely dumb when it comes to networks and servers.

Thanks in advance!

Anyway, thanks again for everything!

Best,

Chris

Glad you liked my writeup.

It is definitely possible to set up multiple servers behind one modem. You should be aware of some restrictions, though.

First, in OS X, you don’t use root accounts, but administrator accounts. They are basically regular users who has the ability to escalate their privileges. For the most part, as long as you’re in the GUI, there is no difference, but ass soon as you are in the shell, there are things you cannot even do as an administrator (looking at other administrators’ files, e.g.). You can circumvent this by using “sudo bash” in the shell, which gived you a “real” root terminal. Don’t do that unless you know what you are doing, though, as you can completely destroy your system by mistake like this.

Second, regarding having multiple computers sharing the same external address. That is possible, but with some restrictions. Let’s assume you have two computers sharing a single public address (so you have an internal IP for each and they use NAT to share a single external addfress). In order to get to a service on a computer, you need to know it’s address (IP) and the port number of the service. Web-servers typically run on port 80 and ssh servers on port 22 (and so on for many pre-defined services). If you type in an address, like westergaard.eu, it is translated to the IP of the correct web-server, which is then automatically contacted at port 80. It is possible to run web-servers for two DNS names running on the same server, using some magic in the HTTP protocol (the client tells the server which one it wants; for example westergaard.eu and lfpedia.com point to the same IP but different sites). You cannot, however have two servers serving separate web-sites from the same external IP. An IP and a port is routed to a particular machine. You can circumvent this by running web-servers on different ports, so you contact, say, www1.westergaard.eu and www2.westergaard.eu:81, where the 81 indicates that the web-server does not run on the standard port 80, but instead on port 81.

You can do something similar with ssh. Simply run the sshd daemon of different ports, say 22 for the “main” server and 2222 for the secondary (port 23 is used for telnet and while this is rarely used anymore, using it for ssh may get you a bunch of unwanted attempts at breaking in via telnet). You can use any port you want and most clients will understand this – use the -p switch for the terminal ssh client. Then set up your firewall/router to forward ports individually based on port number. On my old firewall, part of my service setup is:

where 10.0.0.20 is my server and 10.0.1.2 is my desktop. Here my web-server is sent to my server whereas ssh is sent my desktop. You can also send the same service to different machines, but not using the same external port.

If you prefer not to have different external ports, you can just have one “main” machine and ssh to that. From that machine, you can then reach the others on your internal network.

I have (almost) always gone with the last solution, as remembering weird port names for each machine becomes a pain.

I should say you also have a third, secret, option. It is to internally use IPv6 and have a router that understands one of the IPv6 over IPv4 tunneling protocols (6to4 is the most popular one today). You then have 16 billion billion IP addresses for your machine (from just one external static IPv4 IP). Assign each internal machine an IPv6 address from that prefix and have your machine correctly router 6to4 traffic (all Apple routers do this correctly). This of course requires that each machine you want to use to connect to your internal network had access to IPv6 (natively or via a 6to4 gateway), which is the biggest problem. I used to use this solution (though using an older tunneling mechanism called 6bone), and it works well. It is by far the most advanced one to set up, though.

Finally, you should ash yourself that if you need multiple servers that all should be accessible externally if it is not really worth it getting a real hosting solution with separate IP addresses for each. More often than not, the limiting factor is not the server but the internet connection. My site is hosted at home on a single Mac Mini on a 10/10 Mbit/s connection and my Mini is still also acting as internal server, including eyeTV server, and is running at 10-20% load.

First of all, thanks for posting this. Very helpful.

Second, I wanted to get your thoughts on setting up a couple “remote” Mac Mini servers as I have zero experience setting up servers. Essentially, I have 2 Mac Minis sitting at a friend’s house that share an internet connection through a 4-port DSL modem. From my home, I want to be able to SSH into each of these servers independently, using a root account for each server. A) Is this possible? B) Being that the ISP assigns a single static IP for the DSL connection, how can i also assign an IP address for each server that will resolve over the interwebs when I try to hit them from home via SSH?

Any insights would be awesome – thanks!

I must say, though, that the wiki server is fairly limited, and unless you plan to use it using an OS X directory, I suggest switching to a more powerful wiki server, such as MediaWiki (used by Wikipedia), which is easy to set up and use on OS X (it uses MySQL, which is pre-installed on OS X). You should even (though I’m now just speculating) be able to get MediaWiki to authenticate against a OS X directory, as I’ve previously successfully made MediaWiki integrate with LDAP, which is also used by the OS X directory service. LDAP support requires a separate plug-in for MediaWiki.

Have difficulty finding out how to change am/pm to 24t in the wiki calendar, do you have no idea how to do it?

Sure. It depends a bit on your usage scenario, though. Basically, I can think of three scenarios and they have to be resolved in three different ways.

Scenario 1: You want to use the server on an intranet to send mail outside, but ISP is blocking outgoing traffic on port 25.
This is a common scenario as many viruses set up a mail-server for spamming. To prevent this, the ISP blocks port 25 and forces you to go thru their mail-server. The fix is just to set up the Mac mini to relay thru the mail-server of your ISP. You’ll get an extra header, but your server handle everything. Set up a forwarder like this:

Scenario 2: You want to receive mail for your domain form the internet, but your ISP blocks ingoing SMTP.
This is equally common, even though I cannot see any obvious way to exploit incoming connections. Perhaps it’s easier just blocking in both directions? You cannot fix this without help from your ISP. Some ISPs are friendly enough and set up a forwarding mail server you can use. The idea is that you set your own server with the highest priority and their server with a lower priority. Clients on the internet will not be able to connect directly to your server, and thy the next on the list. The ISP server will then do the same, but is able to connect to your server. If your ISP does this, just set up DNS as here:

If your ISP does not do this, you’ll need an external server which is ready to forward on another port or via SMTPS.

Scenario 3: You want to send mail via your server from the internet (i.e., have a single mail-server no matter where you are).
This is blocked by the same measures as in Scenario 2, but the solution is (can be) another. The easy solution is to enable VPN and just send externally as if you were at home.

The other solution is to set up your server to use SMTPS, i.e., encrypted SMTP over SSL. As this cannot be abused, few providers block this (as outgoing SMTP is needed anyway for this to be abusable). Simply enable/require encrypted SSL and while you’re at it do the same for IMAP/POP. There’s really no reason not to use encrypted connections. SMTP over SSL runs on port 587 or 465, so you’ll need to have these ports open. SMTPS cannot (as far as I know) be used to receive mail in a standard way. Set up encryption here:

Both are huge topics that cannot be easily conveyed. Basically, IPs function s the addresses of computers on the internet (think phone numbers) and DNS functions as the white pages in phone books (mapping names of persons to phone numbers), by translating names such as westergaard.eu to 145.120.13.213 (my current IP address).

For the best understanding, I recommend reading a text book on networking (I’ve used Tannenbaum & Steen: Distributed Systems: Principles and Paradigms. Second Edition from Prentice Hall in my distributed systems course, but there are many others) or using Wikipedia as a starting point (e.g., http://en.wikipedia.org/wiki/Internet_Protocol).