The last of the hurdles to overcome for the deployment was the running of the DNS server. This is because we run on a private IP range internally and use ISA to match external IP's and ports to the services we want to publish (i.e. NAT). This basically allows us to lower the attack surface because we only let out what is needed and an also mix and match servers to the same IP (lowering our IP address usage).
This also means that we have not only DNS servers to allow the servers and staff internally to find the other servers and services but we also have to have external servers too to allow users on the big bad Internet to find them. There is so much duplication of work for this configuration deployment scenario as you are having to create records on a best case of two servers and worst case is four servers and configure them differently. This also means the area for mistake is increased considerably. The upside is that internal staff do not need to go out the LAN and back in via the net or even go through the external firewalls and that we an have different domain names internally and externally, which is great for testing and development and only publishing when needed.
What I do not understand is why the DNS server team at Microsoft can't take a leaf from MSCRM 4.0's IFD deployment and allow you to specify what the internal IP range is and allow you to set A/CNAME’s for both internal IP ranges and external IP ranges. So when an internal IP requests a resolution it gives the internal A/CNAME records and for non-Internal they get the external A/CNAME record. This is such a logical thing to do, that Bind has this feature for ages, so come on Microsoft steal another idea from Linux ;)
One of the design choices for the DNS structure is a concept of mine called IP address abstraction. The idea of DNS is to get us away from IP’s but the problem is that in normal DNS configurations you end up with loads of A records and the moment you need to change IP addresses you end up with spending days changing IP addresses through all the records. What IP address abstraction is that you take a core domain name, and create a single A record for each IP you have.
- internal1.test.com A 192.168.0.1
- internal2.test.com A 192.168.0.2
What you do then is everywhere else you use CNAMEs to those names, regardless of what the domain name.
The advantage is that if the IP’s change ever, you change them in one place and it reflects everywhere, yet the experience to the end user is the exact same as DNS has always been.