Does 2012 OSD get around requirements to get to the MP during the task sequence?

Question

Currently right now I am working in an environment where we are on SCCM 2007, but will be going to 2012 next year. We are in the process of trying to make it so we can build on a secure network that cannot route to any other network. I am prepared to put
a DP out on this same network, but the problem I am running into is that during the task sequence starting, it downloads the task sequence from the MP, not the DP as it is not associated with the package. The MP is not on the same un-routable network. We will
need to have about 6 or so of these, so we do not want to setup 6 MP because now you are looking at not only 6 primary site licenses, but 6 SQL installs (Our environment will not let me get away with doing SQL express for support reasons) which adds up quickly.
Below is a link to exactly what I am speaking about:
http://blogs.technet.com/b/configurationmgr/archive/2009/04/16/configmgr-2007-primary-site-mp-is-used-for-the-task-sequence-even-when-deploying-osd-images-to-secondary-sites.aspx
What I am wondering is if SCCM 2012 OSD was structured differently in a way that access to the MP is no longer necessary during the task sequence process (say for example the task sequence xml is replicated to DPs too)?
Thanks. 

Answer 1

My apologizes. I was meaning to say Primary, Secondary Sites. Not MPs. I brought this up to give reference to idea about the boundaries that I am going to have to use to prevent these clients from contacting DPs they they can't route to.
The idea behind getting off the build network and onto a prod is that security requirements from my companies Infosec require that we limit the amount of traffic on production networks. The idea is that the server be built on the build network, disconnected
and brought into prod. The whole time the MP would be the same, but the DP would switch based on the prod network they were brought into as DP will reside in these prod networks which will get flipped over by the boundary for the prod subnets. Once on the
prod network they would be able to contact the MP. 
I would love to do stand-alone media, this is one of the first roads I started looking down as I have done them in the past for offline builds, but I need the machine to auto assign its hostname from a external servers that will script the build. The build
process we are trying to put into place will be started from an external website. My first thought was to have it add the machines to a collection and add a variable that would be turned into a task sequence variable during the TS and use it for the hostname
when it attempt to apply the Windows settings. The only way I think I could get this to work would be to write a script that is executed during the task sequence that would match the MAC address of the machine and assign the hostname by calling on another
share that is written to by the external website with the MAC and hostnames. The only problem that I cannot get around is that this build process needs to be fully automated and this is for both virtual and physical. I have no problems mounting a stand-alone
build ISO using a script for VMWare, but the problem starts when I look at the physical and limitations of scripting for both iLo and iDRAC for performing this step. iLo can be done, but iDRAC is limited. 
As for a reverse-proxy I will not be able to look at that. They have specific requirements against using reverse-proxies in only specific situations and this it not allowed. I got turned down on this one. This requirement is created by a federal regulation
my company has to follow.