It sounds like the author has a particular argument in mind but is refusing to say what it is.<p>> A popular synonym to what the majority is doing is "the modern way", or sometimes it is even called "best practice", which can be quite misleading because then best practice really becomes stupid practice.<p>Which practice, exactly? The term, <i>best practice</i>, is sometimes used naively but it does have a meaning and refers to the collective body of knowledge we sometimes call, <i>the state of the art</i>. Software has a state of the art. The folks at the IEEE try to maintain a published version of their collection of these practices in a guide called, SWEBOK [0]. If you look you will probably find others.<p>When someone is not being clear and they use the term, "best practice," it's often far too easy to tell if they're simply regurgitating what they've heard from someone else or if they're referring to the state of the art. A few leading questions will often put you on the right course.<p>I don't think this is harmful. It's easy to let someone get their idea across, ask them questions, and if you think they're ignorant then point them to the right place where they can learn more.<p>> Never ever follow or submit to conformity just because that is what the majority or someone popular is doing!<p>I think best practices are important. It's how we push the boundaries of our work and figure out, collectively as an industry, what works and what doesn't. We've known for a long time that unrestricted use of untagged pointers is bad, we have developed a large body of work that supports this case, and research has happened which is enabling us to move forward... with the state of the art. We can use formally verified compilers, separation logic, language tools available in Rust, linear types in Haskell and Idris2 -- nothing is stopping you from using such C-style pointers, but we have better ways to working with them.<p>Some sometimes you totally should do what other people are doing. Especially if they've been doing their homework and publishing their results and having them reviewed by peers. Like the age-old advice: don't write your own crypto. That's a best-practice.<p>[0] <a href="https://www.computer.org/education/bodies-of-knowledge/software-engineering" rel="nofollow">https://www.computer.org/education/bodies-of-knowledge/softw...</a><p><i>Update</i>: forgot to add the link, clarity on state of pointers