I
think it is time to write something more... I am in the process of acclimatising to my new home and getting into the domestic rhythm...
So, I have been continuing to do my research into distributed systems and in the last week I have been doing a lot of reading about gossip protocols and how freakin fast and scalable they are, it has become a major focus of my work at the moment is to soak up the information as well as I can, and synthesise it into a core architecture.
I am really uncertain about the branding, though I of course like my own logo design :) I am very aware of not muddling too much esoterica especially of an identifiably anarchist bent - in the process of my thinking I am trying to see how these solutions apply to a large part of the population - theory is all very well and good but accessible technical information is pretty rare.
My own experience shows that the bitcoin canon will keep you from bumping into the really golden information - that of distributed systems research. Bitcoin has some kind of gossip protocol for distributing transactions, blocks, and its core architecture goes somewhat overboard towards praising the power of entropy, and the upshot of this is the long time for resolving the chain into its canonical form, which mainly means waiting sometimes hours for a transaction to clear.
All blockchains are peer to peer networks, they are built on an architecture of what is called in distributed systems 'replicas', which store the same database across many nodes so that except for the very latest changes in the database, most queries can have answers sent back very quickly. It was notably Xerox who first built a distributed database synchronisation system that used gossip protocols. Gossip protocols are about leveraging network effects to transmit and update replicas to the same state as quickly as possible.
They have other applications too, like maintaining databases of reachable nodes in a cluster - and every blockchain system is a distributed database cluster.
So I am quite deeply stuck in a process of assimilation of a lot of new data and now that I have got so much out of these papers I have seen so far, I am also looking into other things like GPU acceleration of node graph walking - on that subject, the conclusion I have come to is mostly simply that CPU memory is a lot slower than GPU memory (and as I discovered, AMD APU's do not run OpenCL and even if they did it faster the bottleneck is the memory speed.
So as an aside, I predict that the emerging growth area in microchip production is very large cache processors, that are systems-on-die in as far as the requirements for cluster computation. It is no coincidence that very fast memory and optimised pipelines are the key to the ASICs.
Never mind, anyway, Proof of Work is dead in 10 years. Mine it up, it won't last forever. But as Steem has demonstrated, producing creative content, even up to and including computer software, is how people will make money on the internet, and how money will be printed on the internet in the future.
Even if I don't manage to actually implement anything useful...
update
Reading back the bit about memory bandwidth, it reminds me back to thoughts about how proof of work is eventually going to hit up against parallelisation problems and mainly, memory bandwidth.
Nothing travels faster than rumours! Einstein was wrong!
The subject of distributed systems is absolutely mindboggling. How do you say who did what first when you can't trust the accuracy of everyone's clocks on their timestamps?
I haven't quite fully grasped how it's done, but, there is ways to do it. Lamport timestamps and causality discovery protocols... Vectors... I wish i had hated school less than I loved mathematics and programming :(