Formation Packet: Fill & Download for Free

A2A - thanks. Disclaimer: this is speculation...I can't comment directly on any changes, but when we benchmarked a tightly-coded embedded db engine I worked on versus SQLite a few years ago, SQLite was generally slower by quite a bit, indicating that there was a lot of room for performance improvements.As a rule, the thing that kills performance in DB engines (after bad query planning, etc) is copying the data. I once evaluated (old) Postgres, and figured out that it copied data over 10 times between the storage engine and the final prep of data for transmission to client-side APIs. Data was copied on many "module boundaries" between the storage engine, the executor (in several places at the time), projections and aggregations, and final client-side data transmission.The longer you can operate on data directly in native page structures, the faster your engine will be. Ideally, you won't need to copy data until you actually have to format packets or strings to send to front-end APIs. Sometimes, you have to copy data if you're doing a big sort/group-by or similar operation, but a tightly-coded DB engine should be very careful to only copy when absolutely necessary.I'm not a SQLite commiter, but my guess from our benchmarks is they streamlined some internal operations, and probably got rid of some extra copying they were doing.

“Why do some USB ports stop?”As Dave Haynie explained: “Windows, in particular, doesn’t like it when presumed USB devices behave like not-so-USB device.”I can confirm this. Windows does not like devices responding garbage in correctly formatted packets. Either the USB port is paralyzed for a while or till the next reset, or you get a BSOD. We had similar experience with IrDA as well: our device under development, with broken IrDA stack, wirelessly killed any PC within range.

The amount of documentation varies depending on the complexity of the software and the organization of the project team. The kinds of things I'd look for would include:The requirements. Some representation of what the user wants, along with anything derived from those. This would include function and non-functional (quality attribute) requirements.The use cases and actors. This is the who or what that interacts with the software system and what kinds of actions those actors perform. If there are multiple roles of users, those should also be identified here.The interfaces. This is mostly communication and hardware interfaces. If there is a user interface, generic identification of the type of user interface (graphical or textual, web or desktop). If you pass messages, define the message formats - packet structures, XML schemas, text file formats, database schemas.If the software is a distributed system, identify the major components and where they are located. Identify optional components, components that exist as single instances, and so forth.From here, I think you just get into more detail designs. You would start filling in the details of any components and decomposing them into smaller modules, until you end up with the more basic building blocks: classes, methods, attributes, files.