T-Mobile in the UK does this. They also re-write the HTML. This line is added after the first <html> - <script src="http://1.2.3.8/bmi-int-js/bmi.js language="javascript">
Images come from <img src="http://1.2.3.11/bmi/ycombinator.com/images/s.gif
And they add an alt message to tell you how to update the images to the original images. Which is annoying if the alt originally contained useful information. (The original alt comes back if you update the images.)
I'm in the UK. Having images served from 1.2.3.x is suboptimal.
I ask since 1.0.0.0/8 is APNIC and 1.2.3.0/24 is the APNIC "Debogon" project. For notes, a "bogon" is an supposedly unroutable address, or more accurately, an address that you shouldn't see in use.
1.2.3.8, 1.2.3.13, etc. I'm in a coffee shop at the moment, but when I get back I'll try and get a list of the IPs that are used. (The last digit is always quite small though.)
Address space that hasn't been allocated by any of the RIR's (Regional Internet Registries like APNIC, RIPE, ...) are sometimes used without permission, and usually for nefarious purposes. These are also considered bogons since you should never see those addresses in use.
When you see a bogon, something is definitely wrong. It could be your service provider is misusing address space that hasn't been allocated to it, or it could be something far worse (malware, compromised network routers, ...).
The "Debogon Projects" and "Bogon Monitoring" are run by the various RIR's to find those who are squatting on misued address space, and also to get firewall sysadmins to no longer block the unused ranges. Usually, following the allocation lists of the RIR's is sufficient, but some folks don't update their firewall rules as often as they should.
The urls have the same 'bmi' stuff mentioned above, apparently they are using a similar implementation.
When a network/ISP misuses unallocated address space by routing the traffic to something internal, this prevents the inside of the network/ISP from reaching those addresses normally. Unallocated address space can be allocated by the RIR's at any time, so misuse of unallocated address space results in parts of the Internet unreachable.
If some huge networks/ISP's (Comcast, Verizon, Sprint, ...) decided to misuse the address block allocated to you for some internal purpose, you'd be rightfully upset since it would prevent all users of those ISP's from reaching your service/servers. Now let's assume you're a new company and just got a new allocation of addresses from the RIR only to find out the users of major ISP's can't reach your service because the ISP's have already misused your address block for something internal on their networks. Yep, you'd be livid, and livid with good reason.
If you put a lot of work into your misuse of unallocated address space, all that effort could turn out to be wasted a few hours later when the block you misused gets allocated. To reach the newly allocated block, you'd need to redo all that work over again, correctly.
I was mostly trying to confirm that it was the carriers and not malware or whatever.
This is T-Mobile mangling stuff. They do some other things which are annoying, but not Internet breaking. It's a sub-optimal Internet experience, but pretty handy for what I want to do.
What damage could an attacker cause? (And how would a VPN help?)
Many ISP's serve requested pages from massive caches. Although provider caching can improve performance in some cases as well as reduce bandwidth costs for the ISP, it can often result in stale information being passed to the client (you!).
Another common speed/cost improvement for ISP's is serving degraded images from their cache -- by recompressing images (jpg) at a higher compression ratio, the file size is reduced at the expense of degrading the image quality. This is extremely common on mobile networks, but it is becoming more common with land-based providers. In most cases, you'll never notice, since you'll just assume it's a crappy image from the original server. On the other hand, if you do any work with images, then you could be stuffed by the modified images being delivered to you.
Having a consistent endpoint provided by a VPN provider can also be a real advantage. For example, if you're doing checking, testing or trouble shooting, against a system on the `net, then knowing what traffic is yours in the logs can be real helpful. With dynamic IP addressing, your endpoint (public IP) always changes. When you're using a VPN and helping out a friend with something as trivial as reporting bugs, you can tell them that the funky traffic from xxx.example.com is just you running some tests. Even if the site owner isn't your best friend or anyone you really know, it's great when reporting bugs to say, "Hey pg, my traffic is always coming from la.tunnelr.com" so it's easier for them to find it in their logs.
If you need to do your own pen testing across the `net and your ISP does deep packet inspection (DPI) and egress filtering, then once again, you're stuffed without a VPN or unfiltered remote host. Of course, you need to be on good terms with your VPN provider and let them know in advance that you'll be sending some dodgy traffic over their network, but that's not a big deal most of the time.
If you participated in the recent Stripe.com Capture The Flag contest without using a VPN or staged connection (ssh), then you really didn't put very much thought into what could happen if some malicious person rooted the game server and attacked the game participants. Sure, the Stripe folks are fantastic, and they keep an eye on things, but no person has sub-microsecond response times. ;)
Also, some ISP's have bandwidth caps and automated thresholds for reducing connection performance, but they usually have a stipulation in their contract excluding VPN bandwidth from the cap/limit accounting. The reason is simple; business customers would use another ISP if the caps/limits interfered with doing work, and most (sane) businesses provide a company VPN to their employees for remote work.
I use http://www.tunnelr.com almost all of the time simply because it makes my mobile (EVDO VerizonWireless) connection a lot more consistent and reliable on UNIX (OpenBSD). A lot of mobile ISP's don't support UNIX at all, and they expect you to run inane and unaudited software ("VZAccess" which is actually just rebranded stuff from SmithMicro). Having a SSH connection present prevents some of the (intentional) oddities of mobile connections (e.g. "pausing" the link/connection).
There are plenty of good reasons to use a VPN that don't involve bittorrent or similar. The best reason of all is if your ISP does not provide all of the exact details of their filtering and caching methods --of course, none do.