DOS ain't dead

A Plea for Lean Software -- Niklaus Wirth, 1995

Some people (proudly!) have an aversion to "old" things, even if "older" works perfectly fine (or better!). Apple seems to give up on things after seven years. Java seems to deprecate 10+ year-old code. Even Linux now only supports being built with GCC 8 at minimum.

The other bad idea is always "hardware is cheap" (e.g. RAM), but that is clearly not true anymore. RAM prices are going WAY up! I honestly hope this causes some of them to optimize their software more instead of being wasteful. Some people (even Linus himself) say that unused RAM is wasted, but the idea that you can (or should) infinitely add more RAM to x64 is just plain bonkers.

On UNIX, ed would read the whole file into memory. Thus, sed was born to only read a line at a time (for bigger files). Gone are the days when slim software was valuable. Gone are the days of trying to cram something useful into CP/M (or DOS 640k or IA-32's 4 GB of RAM). Firefox doesn't even make 32-bit builds on Linux anymore.

I'm not really raging, but it's sad how bloated we've gotten. Optimization takes time, and some people just aren't willing to do it.

> Give me more fallacies to debunk!

You should probably add security, to your list. But not without providing your own insight, on it.

Something that is not a fallacy, is that software bloat can be a mess to maintain. Bloat, and complexity of maintenance, provide corporate advantages; especially in the face of the opensource code. I question, if this isn't the real reason for interest in Rust. But that really depends on what features of the language your are using. If you are aiming from extreme security features (as it is claimed anyway) that code will be harder to maintain and harder to fork (hard directional design forks). However, one of the reasons for pushing Rust, might be the idea that AI code could need less auditing. The idea of saving the money, because time earned experience is obsolete.

Despite having no way to mitigate some of the notable CPU ?flaws?, how susceptible is DOS to them? This isn't a fair comparison, due to the limited usability "of DOS" in the roll of a server. But, that is were you begin to unravel the real issue. These modern approaches, themselves, are the security nightmare.

There is an existing disgust, for operating systems running in ring-0, if connected to the Internet. I really more prefer the idea that ring-0 privileges come from direct hardware keyboard access. From there on out, what you do is on you. But instead, the secure idea is that ring-0 is attained from dynamic privilege access. There are practical reasons for this, but it doesn't change the inherent nightmare that comes with that. No matter how responsible you are, with the applications you configure and run, your kernel is intentionally built with hard coded backdoors (role-the-dice, or maybe even less secure than that).

The idea used to be, protect the system from bad code, that might compromise system integrity/stability. This was added, at a hardware level, and has pros and cons. I wonder what an OS designed for unreal mode might be like. The amount of code included to protect the system from userspace is crazy. Everything you add, is itself a potential vulnerability; and must be maintained to match future secure code modifications. Add to this, that now, security must be considered everywhere you look.

There are practical reasons for these modern designs, even though I disagree with them. I really consider computers a tool, not something that the infrastructure/security of the world should depend on. If we are reaching a point where computers are the only option, we are doing it wrong.

The elephant in the room is the aging idea of "personal computing". Nothing modern is indented to provide a person with a "personal computer". You are buying a social device. I don't mean "person to person" socializing, though that does play a large role. More and more, the software (and somewhat the hardware) is being designed with usability that highly merits network/Internet access. Some devices are near worthless, without Internet access.

All of this creates a complex and hairy ecosystem, that adds convenience, at an extreme cost to simplicity. All of this complexity is fine, if you just "go along, to get along". But, that is not "personal computing". You must move with the herd, and that can mean both software and hardware.

Software wise, we all took the convenience bait, even if we still prefer simplicity. The consequence there is that, the development of true alternatives fell away/behind. Probably the biggest cause is that, there wasn't a big enough market for it. But, corporate interests have the upper-hand, with greater control over available hardware. While that advantage has been abused, realistically, it could have been abused more heavily; and it is about to be.

What will remain of personal computing, when the computer you use must be authenticated to your legal identity; and any alteration of the device, or its software, voids its ability to connect to the Internet, is illegal, or causes the device to break upon detection? Some will have old devices, an those that we manage to modify, for non networked use. Will we even be able to distribute software "online" for non-secured hardware, and operating systems? If the software is for a modded device, it might be illegal. It my seem extreme, but older operating systems and hardware might be considered security threats. That might be a long ways out; but its approach could be greatly accelerated by worldly catastrophic events, of a digital security kind. Hmm, not that it looks inevitable or anything.... If not for something dramatic, we have had a pretty good track record, of moving away from things really slow. Maybe that will help.

The biggest fallacy, that supports software bloat, is that we should "go along, to get along".

> New computers are more efficient than old ones;

I think they are (the HW), if we measure computing power of modern CPUs MIPS per watt they are definitelly better than old ones. And now you can choose from huge scale of computing power from very low consumption Quark, Vortex, ULV Atoms and i3 to huge multicore monsters. But the problem is that bloatware demands more computing powet for the same tasks than before so the HW progress is neutralised, sometimes... I can imagine one writing a letter in MS word office 95 at windows 95 running on some Pentium 100 32MB RAM, 30 years later one doing the same in Office 365 on some 64-bit corpo laptop i5 4GHz 16GB RAM, did the productivity of writting changed reflecting the computing scale? Maybe now with AI, now it writes letter for you based od simple request :)

---
DOS gives me freedom to unlimited HW access.