This is just a little sketch of an idea at the moment, but I think it is enough. I might write it into my development roadmap for steemportal. Basically it is a Steem Witness Proxy node.
This thing will accept all the standard RPC queries that a full node would, but it only keeps data in a cache that has actually been requested, and it will download query results from multiple random nodes to verify and protect against Sybil attacks. It would then also make sense for it to keep score rankings for divergent results and deprioritise their results and frequency of queries with a slowly regrowth in case the particular node stops doing this.
It's like a web accelerator proxy, for Steem. It would allow for smaller platforms to run such a proxy locally, and accelerate the reloading of old content. It would also limit bandwidth usage as well. It is a better solution than the relatively centralised clusters used by electrum for bitcoin. It can be integrated into any steem-using application as a minimalistic version of a full node, that just stores relevant data to a user, without having to trust the network, by ensuring data consistency between its sources.
It would allow lightweight Steem applications to be built where a user has either storage or bandwidth limitations. I think eventually there could be a payment system where you pay the RPC node operators for blocks of data transfer as well. Seeing as this does not accrue to any other relationship with the nodes, it does not compromise identity, but merely rewards RPC node operators for their services.
Protecting source address is an issue, and could be solved by using a Tor tunnel, indeed, Witness RPC operators could also use them to keep the traffic inside the tunnel. But this is not so important, since it is just an additional layer in the transport. Actually, this could be a way to move towards making it profitable to run Tor-like networks. Witness node operators could run both the hidden service and an internal relay only node. Some of the profits from paying for Witness RPC nodes could contribute towards running such relays, the first instance of which I can see a way to make money running a Tor relay.
