Totally agree. I think you’ve picked up on an attitude problem I need to fix, as that is keeping me from embracing a really useful technology. You caught me admitting to a bias that I know isn’t always true.

I’m a bad one to get how-to advice from if you’re starting out. Not a fan of docker and I don’t know what watchtower is. I’m one of those electricity-wasting home labbers who loves ESXi, vlans, and /30 nets for each individual VM.

I’m also one of those who takes months to accomplish what someone competent can do in days. It’s taking me forever to get openldap, postfix, dovecot, and roundcube to all play nice. (Because I’m trying to “be like daddy” and mimic the security I see at work, I can’t follow normal walkthroughs, or just install an off the shelf container and make it someone else’s problem. But this way makes me read manuals and gain a deep, durable understanding of the technology. And it takes forever.)

Some browsers have an option for DNS over https, and might be skipping your system DNS settings. Other advice is right too: run this locally and don’t expose it to public internet queries.

Security is a tough thing to give advice about. Different people have different levels of risk tolerance. It’s embarrassing to give advice about one’s personal views - tedious to write - and then get replies about how that’s too much security, too little security, etc.

Attackers can use tricks to enumerate dns subdomains. They can compromise one container and pivot to the container host.

You can frustrate automated compromises by putting up roadblocks or speed bumps they have to get through before seeing the stock landing or login pages for well known apps. That can buy you a little time if a serious exploit is discovered and you know you won’t be on top of container updates. But stay on your container updates.

Yep, Longest Common Subsequence is usually greedy and that’s the earliest set of lines that satisfies the search. Happens when you just treat a file as lines and only match those.

You can get better results with more syntax or content awareness. Chunk into paragraphs or code blocks or functions, then sentences or statement lists, then lines, then words, etc. I think Beyond Compare can do this.

Current implementation seems to focus on administrative domains for control, like email servers with individual policies and reputations. What if we look at this the other way?

People have different value systems. Are you ok with promotion for monetary gain? (No never / only individual contributors promoting themselves / only small businesses and below / yes) Are you annoyed by $controversial_topic? Do you dislike when bored people make a conversation game out of someone else’s need for obscure technical help?

The details can be decided later by people smarter than me. The point, though, is that these value systems aren’t universal. Users should decide their own.

Meta interactions (up down report friend block) should be aligned to these values. My client would gather meta-mod data as well as votes/comments. I could easily configure my client to hide things, or group similar distractions together and show/hide them all together. Your client could work differently.

I have no idea how we would possibly implement this with federation. Civically minded users create a meta-moderation identity with a PGP key, sign and publish their decisions, and let people choose to trust them based on past behavior?

Probably still flawed, susceptible to karma farming and cashing out. If well known mods start betraying their users, the bad activities are signed and can be used as proof they can no longer be trusted, though it could take days to get people to stop trusting someone.

Even the whole value system idea can be subverted. Dog whistles, toxic in-jokes, things which are offensive in context but seem fine judged later out of context, etc.

But I want this for us all. (And I vaguely remember seeing something similar on slashdot in the 90s) I have no idea if Lemmy can even support it though.