Perhaps for small networks it's trivial. This was a series of four medical offices which merged. I was part of the effort to standardize the apps and network. Including remote users and VPN users, there were about 200 machines, which was barely enough to be accommodated by a single Class C IP block. Instead, we renumbered the networks of each remote office to avoid duplication problems. It was not particularly complicated, but it was time consuming and required a rehearsal.
For a typical home user, this problem is not even an issue. The IT department of wherever they're trying to VPN to takes care of the issues. In most cases, they simply force ALL traffic from the connecting client machine to go through the tunnel. The disconnects the user from his own LAN, and forces plenty of wasted traffic going through the VPN, but is the safest and most secure method. One could even have duplicated gateway IP's and it would still work. Too bad local network printers won't print, but there are workarounds (i.e. USB printing).
Sure. Let's try a bit of math. I have about 15 customer sites that run a VPN of some manner. Each site consumes about 10 static IP's, 10 dynamic IP's, and needs a VPN IP pool of perhaps 20 IP's. That's 40 IP's per site. If all these sites subscribed to the consumer brand of VPN router, which defaults to 192.168.1.xxx, I would need 15*40=600 unique IP addresses to avoid duplication. Obviously, this is not going to fit in a single Class C IP block, which allows only 256 addresses. I can widen the netmask to perhaps /22 for 1024 addresses, but many cheap routers don't work well with more that 256 IP's. Obviously, not all 15 customer sites need a tunnel between them, so this estimate is worst case. Still, it does illustrate why I have a simple rule for assigning IP blocks for remote sites with VPN's. I pick a random IP block starting with 192.168.[3-254].xxx. I avoid building networks using 192.168.[0-2].xxx as these are where the typical home routers are located.
Incidentally, since I started doing this perhaps 10 years ago, I haven't had many address conflict problems. I've also fixed a few small networks that were having weird call in problems by renumbering the office LAN so that the home users can use whatever IP block their router manufacturer finds fashionable.
Agreed. I think I stated that when I mentioned the problem of duplicated gateway (default route) IP addresses.
Of course. Many times with many variations on weirdness and failure depending on flavor (PPTP or IPSec), hardware, firmware, client software, and versions. I have the resources, but I'm lazy/busy and don't want to do anything more while waiting the worlds slowest backup to finish (USB 1.1).
This is interesting but I think we're way off the original question and subject, whatever they might be.