It's all well and good for you to stick with your autistic mid 90s time capsule, but unless your software can exist in a vacuum, you're going to be adapting to ever-changing standards and interfaces.
Hence software should be designed to exist in a vacuum. There's no good reason anything but the highest layer should need to be changed occasionally, if that. Software can be finished.
While your Common Lisp bible is still accurate, good luck sequestering yourself in the mountains and producing anything worth a damn if that's all you have.
It's time to drink.
Unless you want to be like Terry A. Davis and write literally everything yourself.
I do, and that means several things: Firstly, I won't be hit by a dependency getting some chink maintainer who tries to fuck me over; secondly, if my software be flawed, it will be flawed in different ways and therefore less attackable; thirdly and lastly, this means I'll simply never write certain kinds of software, even if I'd like to do so.
Let's pretend I needed to write something using the xz compression format. I hate this trend of software written in superior languages using C language libraries, because the flaws bubble up, so I'd have to implement liblzma by myself, which would take a while, given I'd need to study the format to design the interface, and I'd probably lose interest. The easiest option would be to remove the stupid fucking xz dependency, wouldn't it? Complicated software is only growing more complicated, and this is what large corporations want, because it removes the power of an individual to implement such software alone, with the WWW as the best example.
The trend of shoving TLS into everything is another good example. How many TLS libraries are there? Not many are. Several years back, I downloaded and played with DRAKMA, a Common Lisp HTTP client library. It was fun, and I could do some neat things with the restart system, but auditing DRAKMA and all of its dependencies was a bother, and it used OpenSSL, making it vulnerable to the same horseshit as all of the other programs written in inferior languages. The easiest thing to do is to write a Common Lisp program to generate input for GNU Wget, which can be sequestered across a process boundary at the least.
So, by writing things alone, I distinctly feel each dependency, and can carefully judge which are truly necessary and which are superfluous. Many are superfluous. A fucking SHA-3 library written in the C language had a damning flaw, and it may have been there for years purposefully. I've implemented the SHA-1 and SHA-2 family of functions by myself, which means I'm not affected by flaws in common libraries; SHA-3 is a bit harder to implement, and I have no need for it, so I haven't done it yet, and may never get around to doing it.
All programming languages and all associated packages should have man pages, "man perlfunc" etc.
GNU Info is better.
attention all boomer programmers: please attend todays cobol code review
I have some COBOL books by Tyler Welburn I need to finish reading. It's a fascinating language, and lacks some of the flaws that plague modern programs, believe it or not. Most COBOL is written as a finite state machine, which immediately removes the possibility for several kinds of errors. Sure, COBOL programs are often gated to get good inputs only; gee, is that cheating or a good idea?
