Hypercore, GNUnet, SPA, Optane, A100, TCP, Consensus, Thunar, Spy Wars
Hypercore-protocol (@ hypercore-protocol.org) and Hyperdrive v10 (@ blog.hypercore-protocol.org). Yet another distributed p2p file system sharing protocol. Let's dig in for details and the good and bad stuff. Sparse downloading isn't anything new. Kademlia / DHT, sure, nothing new. Mounting network resources as "magic drive", well, nothing new. Interesting, I didn't expect that the deduplication wouldn't work yet. Freenet and GNUnet both solved that long time ago, by storing data, as simply data. Which means it doesn't matter which reference originally was used to reach it. This also means that the storage is most likely less distributed than it's with Freenet or GNUnet. Yet because resources are required to find the potential sources, high distribution probably makes over all data fetch latency worse. As example BitTorrent doesn't use high distribution solution. Yet if they'll end up with content-addressed block storage, with quite small blocks, then they're doing it exactly like the Freenet / GNUnet. Yet there's generic question if that weakens privacy or not. New solutions to distributing data efficiently and in a distributed fashion are always welcome. I just wish that there would be more working solutions, and less new solutions. All other solutions suck, let's build a new one. That's an easy trap to walk into. Yet it still can be an amazing opportunity to lean and test stuff. kw: Hyperdrive, Hypercore, Hyperswarm, Hypertrie, Torrent, Dat, IPFS, Tahoe-LAFS, GlusterFS, Reinventing the wheel
Single packet authorization (SPA) (@ Wikipedia) and port knocking. Nothing new. If if server is providing any public services, like a web-server, then that's the simplest way and port knocking / SPA isn't required anymore.
Studied Intel Optane DC persistent memory and it's features. Just to know emerging and latest DC technologies. It seems that there's huge industry on data analytics, but that's limited to specific companies doing the actual analytics.
Checked out Nvidia A100 "GPU" and it's DGX and Ampera presentation video. More hardware for data analytics with well integrated software suite. Is that actually GPU at all? Are we going to Babbage Analytical Engine (@ Wikipedia)?
Read long "What every developer should know about TCP" discussion. It's just like anything else, there's so much disinformation and misunderstanding there. Yet most of the serious issues got fixed in the discussion, but aww, it was painful. After all discussion it came to the normal conclusion. If you don't know and understand how things work, you're going to fail. This is no exception. "TCP/IP Illustrated". Early discussion was about basic flaws, end of discussion was about congestion control, buffer bloat, dead connection detection (keepalive) NAT-issues, etc (LEDBAT, BBR, ECN, etc).
Paxos vs Raft: Have we reached consensus on distributed consensus? (@ arxiv.org) - A really good question. Nobody mentioned Stellar (@ medium.com). Different protocols come to pros and cons pretty quickly, as well as several parameters can be optimized based on the situation the system requirements. Using different compromises and trade-offs. Can nodes be trusted, and so on. For fun, read: "Hundred Impossibility Proofs for Distributed Computing" kw: Quorum, Multi-Paxos, EPaxos, Consensus Algorithms
Found so classic fail in Thunar's progress bar. If copying fails, and you'll need to click retry. Guess if the progress bar is reset? Nope. You'll make it 100%+ percents. And the completion time estimate also flips, probably due to bounds checking missing. When the amount of remaining is negative, the amount of time gets huge, as expected. So classic (bad) code.
Something different? Watched Damian Lewis Spy Wars Series 1 (@ IMDB). Yet I knew all the stories inside out, before this documentary series. Because these are the classic stories chosen for the series. I'm just wondering if the Darknet Diaries (@ darknetdiaries.com) could turn into TV-series.