I think that it will be increasing more difficult to run a validation node at home. Either due to latency as Adrian highlighted in the “Validatorians” chat, or because of the increasing number of technical requirements it will be hard to be a hobbyist validator. Running a validator node as a single person might even be tough.
Even running a validator node in the cloud could be difficult. I suggested this in the collaborative document that I started on validator security.
I totally agree with you, security is the most important part of a validator work, and for the power of the machines, I do not think that the validation process will be very greedy in performance. (thanks for the document)
From what I’ve read, seems a lot depends on the particular PoS, its rules and organization. PoS projects that want to have lots of validators can get away with soft slashing rules as individual validator uptime and integrity isn’t so crucial. Automation and structuring can remove risks and workload - but the software all these validators use must manage things a lot and be safe or whole network is at risk. This would make validating more accessible to those with less know-how and resources. Though, as you say, internet connection latency matters (eg. residential connections = poor) and many validators slows down the gathering of consensus = less TPS.
Where a lower number of validators are wanted then rules will need to be stricter, requirements higher and complexity will increase as individual validators matter more… Ie. only specialized businesses operations could manage to do it.
Yes, running validators solely from cloud would be difficult for both non-pros and commercials due to problems of using hardware key security.
With Polkadot, we hope to at least make the minimum barrier to entry as low as possible but perhaps it will still require some level of expertise that’s beyond the grasp of hobbyists. It is certainly possible that we could have multiple tiers of validators who control the different tiers of relay chains.
The top level relay chain will probably require a high level of dedication to keep the nodes running and secure. While lower level tiers may not have such strict requirements.
I guess there’s other compromises with having a lot of validators besides that latency might slow down the TPS. But regarding latency, tps and the main relay chain in polkadot: Correct me if I’m wrong, but Polkadot isn’t principally scaling by having a high number of tps on the first order relay chain? Instead, Polkadot seems to be designed to scale with a “tree-like” structure of parachains and 2nd-order relay chains. By this architecture even 100tps on the first order relay chain should be sufficient enough in offering more or less infinite scalability with fast finality. This would imply that having a super short validator latency on the first order relay chain is not that crucial, which allows for a higher number of validators. On the other hand, I can see that some niche parachains with specific use cases would need a lot more than 100tps, e.g a game or a DEX. In those cases, latency is much more important and that would effectively limit the amount of validators/collators.
It also feels somewhat counter-intuitive to have more validators on parachains or lower tier relay chains than on the top level relay chain.