To begin with, I do not know this, but the domain is the logical choice.
When restricting IP, multiple clients on the same web server will consume each other's quotas, which they should have thought about.
* The ip client will be useless in every metric. * The ip server will mean that several clients on the same host will consume each other's quotas.
The domain remains. However, with that said, Google knows how to use their brains, and I won’t be surprised if they have a combination to find abuse. Thus.
If domainA.com uses up to 25,000 in one day, and immediately, domain B.com enters the network and starts requesting images from the same ip, which can make some calls.
Of course, the same would be true even for different ip if they all request the same location.
So, I think that if you make a case when a domain requests a card for any given request and only locally mark the domain as spent (for a day), when you get an error, I think you can request an infinite amount (if you endless domains). With a possible warning of discovery if everyone is requesting the same location.
Of course, distributing different domains on different / ip servers will make it impossible to detect, however unlikely it would be.
Rickard liljeberg
source share