Protracted discussion (and promotion) has glossed over one key point:
None of the people on-site are technical, and all their data is accessed
via RDP on a server in the United States.
They will not be happy with VSAT latency (typically 700ms though
physics says you can never do better than 550, and that's for the
space segment alone) if they are running RDP, VNC, Citrix, or similar
technologies. Sorry for being a buzzkill, Warren.
Actually, Citrix (in particular) works quite well over satellite latencies. The network project I'm working on right now is wrapping up an app rollout to about 100 countries, many of which we can only reach via VSAT. Testing showed that Citrix performance is much better for AJAX-y web apps than pure HTTP.
Citrix has a bunch of intelligence built-in specifically to deal with issues related to high-latency/low-bandwidth circuits, including local mouse, local echo, click/movement aggregation into large packets, and of course compression-before-encryption. It's not quite Memorex, but it's very usable.
I'd be happy to share the data with anyone who is interested; I also showed that F5 Big-IP load balancers can make really horrible ERP apps (*ahem* Oracle eBusiness *cough* *cough* *cough*) work a lot better as well, if you decide not to use Citrix.