• 0 Posts
  • 16 Comments
Joined 1 year ago
cake
Cake day: July 15th, 2023

help-circle
  • The concern for code duplication is valid, but as the article mentioned it is also a while off until the Nova project is mainlined. I honestly never thought of how the work to bring in Rust to mainline may in effect lead to a more complete deprecation of older hardware as we start to change API’s older/unmaintained components aren’t updated. On the flip side, trimming out older stuff might save maintainer work going forward.




  • The constraint on memory isn’t on the compiler it is in the available ram and flash on ultra-lowcost/power microcontrollers, use can find some with less than 1KByte of ram, so that 32bit int could have been 4 8bit ints in your electric toothbrush. The packing of bits is space effecient, but not compute effecient and takes several extra clock cycles to finish when run on an 8bit microcontroller. It’s better to store it in the native 8bit word size for these devices. Further more, on more powerful systems using the smaller size can allow you to optimize for SIMD instructions and actually do multiple operations per clock beyond what you could do with the 32bit size.

    There are reasons for these types to exist; as was mentioned elsewhere, if you don’t care you can always just use i32 or i64 for your code and be done with it.


  • Soooo, an int in most architectures is actually signed, usually a 2’s compliment signed 32 bit value; but the spec does not require it to be 32bits, some platforms might use an 8 bit or 16bit value instead (think older 8bit microcontrollers). That’s why they have ‘int32_t’, ‘uint32_t’, etc for C/C++, it just sounds like you haven’t used these number types (check stdint.h). Rust just requires you to use the more explicit number format by default (which I personally prefer because I have had to jump between embedded and Linux development).

    The multiple string types are annoying at first, but its probably better to realize that they are more like two types (String and str) with the apperstand (&) ones being references (think pointer in C/C++). A String is like a C++ std::string and a str is more like a C-String (fixed sized array ish of chars).


  • Enter embedded programming, believe it or not there is a ton of low level C code being written; also, try adding a new protocol for networking, there are many cases where bitstructure matters, I recently wrote a small bit of code for a project that used bit packing so that we could fit inside of a single Ethernet frame so that we didn’t have to deal with fragmentation and the networking overhead it caused.

    For context, what is your past programming background and what are you trying to do? While rust is a great language, it may not be the right tool for what you’re trying to do if these are things that you view as unnecessary


  • A reason to not distribute a library as a binary is for cross platform support, this allows a library to technically be compiled and used in some other platform; with that said, there are assumptions about having the same system calls (or that the library only handles abstract things that don’t hook into a specific architecture), with that said, making cargo smarter and adding in the option to bundle some prebuilt libraries to be pulled for compatible platforms could speed some compile time (at the likely cost of build size, etc).

    Distributing a library as source isn’t some weird rust paradigm, take BOOST for example, which is a large library for C++, and is usually distributed as source (but Linux distros may bundle prebuilt binaries; it is generally intended that you will compile it and use it as a shared library), actually a lot of compiled languages distribute their libraries as source, it also allows you to read their source if you need to see exactly how they’re doing something). I don’t believe there is anything preventing someone from using shared libraries in rust, although they do seem to be less common. Another reason to distribute a library as source is for licensing.

    The unstable versioning… Ya, that is definitely annoying, and likely tied to the limited resources available for sustainment for various projects. With that said, compared to C/C++ many of these libraries are something that one might role their own or use in house libraries. Now for Tokio… You don’t have to use it for async, although you probably will want to (see: https://blog.poespas.me/posts/2024/05/24/rust-async-std-and-tokio-comparison/), I do wish there was a nice way to depend on a library and substitute some of its dependencies (aka Tokio); with that said, part of the cargo idea is that you don’t view these libraries as part of your source code for review (although, auditors would likely disagree, so acrediting with the inclusion of many in accredited libraries would probably not be fun)








  • I usually wear deodorant (which generally has an antiperspirant in it), but there are days when I don’t/haven’t; in this limited experience I’ve noticed I would generally smell worse if I haven’t been active. It seems like being up and moving around and sweating from ‘labor’ rather than sitting and sweating (not sure how to word that) is less likely to significantly smell. With all that said, I do try to shower after the gym or when I return inside after working on significant outdoor projects.

    P.S., another data point to consider is the individual’s age, as hormone levels (generally tied to age) could also influence perspiration


  • It sounds like git-annex to me (doesn’t have to be 3rd party hosts, you can combine your own systems) where it allows decentralized file management across all your devices. I’d also be curious to see how they handle offline backup/file tracking and the sneakernet use cases (something git-annex supports). As it sounds like it is a true filesystem, it could be a nice upgrade assuming feature parity (one gotcha with git-annex, is since it is using git, you can’t easily have other git repositories inside of it).