We have been on fire this last week!
However, I am still trying to figure out why we historically get so few transaction fees. I look back at the blocks we solve and the blocks surrounding ours are almost always >900kB. The last block we solved, block 435656, was only 20kB. I am betting kano is going to jump in on this and say something like, "People do not configure their setup properly. That is the problem you will always run into with p2pool." In any case, I think we should attempt to improve this somehow. Any ideas?
You're spot on with your reasoning. Knowing how to set things up properly plays a large role in the composition of the blocks you create.
In a more typical pool model (like mine, kano's, etc) a single source provides the workload for all miners. You as the miner don't have to know anything other than how to configure your hardware to point to the pool. Details of block size, block composition, etc are all managed by the pool software and the pool operator. With p2pool, when you are running your own node, you must also learn how the blocks are created. Ideally, you should understand the code - how it creates blocks, how often it changes the block composition, etc. For example, what happens in the p2pool code when a block is found on the network? How does it invalidate the transactions on which miners are currently hashing? Does it produce an empty shell and push that to miners? Does it fully validate the new block? What are the optimal block size settings that will be most efficient for your p2pool node? What kinds of transactions should be in any blocks it creates?
These are just some of the things a pool operator needs to know, regardless of whether or not that operator runs a typical pool like myself, or a p2pool node. The reality is some pool operators don't have the foggiest. This applies to both models - I'm not singling out p2pool here.
In a nutshell, on a pool like mine or kano's, the burden of this knowledge is ours to bear. We tune our pools as best we can to produce full blocks with lots of transactions and fees. For example, kano ensures that all of his own pool's transactions are included in the work given to his miners. Why? Because it allows him to confirm the pool's miner payout transactions, which might otherwise not be picked up by other pools.
