Just now read the article and wanted to make a comment. Particularly in regards to what you'd do differently. Doing all of this in memory kind of defeats the purpose of having stored data. It adds speed, that is true, but there are problems in regards to having truly decentralized applications and to have them be verifiable. On the other hand, the way Bitshares operates it would likely not be an issue mainly due to the way that the consensus algo operates. As for the merkle tri-graph, I would like to hear why it is redundant. It is very much expensive, that is true, but if you want a truly decentralized and trustless application, it is somewhat necessary. But again, I can see why you'd do it this way due to the Bitshares algo as the goal in the algo is not to be trustless, just to be decentralized. There are tradeoffs to everything. In any case. Hope all is well.