Learn Node Based on Realistic Limitations

[Ryu945]

I was talking with someone and they told me how in areas with 600 plus nodes, 33% of the traffic is just nodes constantly trying to fill the 80ish slots that are on a device.

https://cdn.discordapp.com/attachments/1161490210165628998/1426638794391818331/image.png?ex=68ebf46a&is=68eaa2ea&hm=b68f32866f226b6934928afe946a759f6190b5ffcf8555f4caaa32597d72192a&

There should be hop limits for learning new devices imposed based on how full a devices node list is currently set. The minimum setting would be to only learn nodes that are 1 hop and 0 hop nodes. It will automatically go up to the max of 6 hops away if there is room on the list.

If a device is at its node limit sees a new device and it is outside of the hop range then it doesn't even try to learn that device. Devices will keep a reserve space of for learning new nodes. To learn a node outside of the automatically set hop limit, you have to manually request the information from a device inside of your hop limit. It will this pass on one by one the nodes it knows within its own hop limit. This should stop the constant data churn of learning new nodes.

For example, let's say a device can only learn 85 nodes. It can keep a reserve space of five nodes. If it has no favorite nodes then it will learn all the nodes it can up to however many nodes fills 80 slots. If knowing up to six hops requires 81 slots then it will drop down to five hops and purge all the nodes that are part of the six hop range. If you add 10 favorite nodes, then the device now only has 70 slots to work with. If it can't keep all the non favorite 5 hop nodes in those 70 slots then it will purge the database of all 5 hop nodes and only learn up to 4 hop nodes.

The reserved 5 hop space is used to manually request nodes that are further away then the current hop limit. You can ask a node within your own hop limit to send 5 nodes that it knows. You can look through them and save any you want to save. If you don't want them then none will be saved. You can go 5 nodes at a time until you look through their entire node list. All of this is manual.

This would 100% stop the constant data churn of nodes relearning each other.

[Ryu945]

Thoughts?