bryarsdj

Or the opposite, I imagine a name like "John Smith" couldn't have this issue because there would be so many matches.

But shouldn't the site owner pay for a CDN and host the resources themselves? In which case the CDN wouldn't own the IP information. I think the problem here is that the website author is getting free bandwidth in exchange for their user's IP address, which in the example Google can then use for tracking and other things in exchange.

Yes, but I think UDP is still a good counter example for this analogy. Buffers can be detrimental to udp traffic for real-time use cases (eg VOIP). Here your trade off is quality (in voice this is jitter) versus latency. Typically you want to make this choice of buffer size at the receiving end only, and keep buffers elsewhere as small as possible - routers which store and forward, network card buffers,OS buffers etc, can add up to lots of bloat.

So for VOIP "JIT" is a good thing and your "Inventory" levels need to be tuned at the receiver.