Anyone have an idea what percentage of networks and hosting providers drop spoofed packets originating from inside their networks? What are the downsides of dropping them?
You're referring to BCP-38, aka RFC 2827. It's actually decently hard for a transit network to do this at scale. You can do it when you're a small ISP, but the administrative stuff becomes harder as you get larger and are constantly getting more allocations, buy out other ISP's, etc.
Every content network should do it. Not a huge win there, but it's something.
I'm not saying we shouldn't try, but there are countless, very long threads on NANOG about why some transit networks just can't do it.
