Mark Tomczak @mtomczak@qoto.org

@rob (Some details changed to protect the actual implementation, and definitely not because it's been years and I forgot them. ;) ).

Once upon a time, Google had a more-than-a-few-minutes significant outage because someone pushed a configuration file for the outer-layer web traffic handlers that included an empty array as the set of valid handlers.

At the time, the system would cheerfully take that input to mean that every single frontend server was in maintenance mode and should hand off 100% of its traffic to other available servers. Google.com requests slowed to a trickle as servers began thrashing, desperately searching for a peer that wasn't in maintenance mode, just sloshing requests back and forth like a bubble bath. Damage was, fortunately, mitigated by the rollout being slow (whether that was intentional for self-protection or an inefficiency that nobody ever got around to fixing, I can't remember).

Anyway, that file was valid but they narrowed the definition of "valid" to "the set of active handlers should be more than zero." I can't remember what they did about the rare circumstance where it was *correct* to say none were valid though.

"So I've got this act. I get a bunch of artists together, they make art. They publish it online in a ubiquitous communications platform that makes it trivial to see and copy their art, but the law says you can't **copy** it without some limitations so they figure they're fine.

So then a bunch a' eggheads, they come in, wave their dicks around, they build a machine that makes new art based on art someone made that gets fed into it. Then **they put that machine in front of the public.** The public puts all the artists outta work creating just the filthiest things imaginable, like they tell the machine to draw "George W. Bush fucking a bomb as the bomb is falling on Iraq, only Iraq is Saddam Hussein with a huge vagina." Just terrifying stuff. And now everyone hates the pictures and hates the eggheads and the artists hate the eggheads and the pictures and themselves and most of all, **everyone hates art.**"

"Well that's a hell of an act. So what do you call it?"

https://labs.openai.com/s/adcpxMnv9UfjVaI4SPEsjYPV?fbclid=IwAR0geEv0HFv_pU5UPLBtaPKsN_NFeKFv9rNJQo8hrkK-9Q5uIEy6WH3oTJI

@jhertzli Thank you!

I think a broad statement is possible on people's intuition of how loaded structural steel works under heat being bad because they think of "melting" as what ice does, and what a conductor does (especially a heterogeneous conductor 1,200 feet long under a side-load, or 1,700 feet high under a vertical load) is an entirely different sort of behavior.

Today I read where someone described the human condition of being the place where the falling angel meets the rising ape "Spicy qualities" and I can't stop thinking about it.

@aberrant68 They appear to come in pairs consistently. I'm leaning towards the other suggestions that they should be interpreted as couplets of five symbols each (so 25 possible values per pair).

@lauren Yes, qoto does delete-repost.

I've assumed the admin thinks mutable records make for dirty pre-temporal responses, but I didn't really dig into it.

@lauren The biggest concern I have about Mastodon in the long run is that one of its strengths is one of its weaknesses.

The protocol is a protocol, but different nodes can implement the protocol differently. So there's a matrix of features one may or may not have depending on what node one is on. For example, over here. I can't just sub to a hashtag the way a quick Google suggests I should be able to; the admin unimplemented that in favor of a more elaborate regex-matching tool... Reached from an entirely different section of the UI.

That kind of fracturing leads to analysis-paralysis and is poison for adoption, but on the flip side how *cool* is it that my subscription is better-than-average? Like, neat!

@lauren This is why I heavily engage in me-ism for 90% of the software I write:

* I don't intend it to be understood by the elderly

* I don't intend it to be understood by the colorblind

* I don't intend it to be understood by anyone but me

* Also, I'm not releasing it

* Also, I will forget how it works the minute I put it down ;)

@togetherwecode Bold of you to assume we work at companies where our decisions have consequences. ;)

... no, more seriously though: there's some wisdom here. If the company is too big, it may take a significant fraction of your career to get to a place where you have enough agency to make decisions with consequences (as opposed to "How do I implement this function my boss told me to implement?"). If the company is too small, engineering decisions take a back seat to issues like market unpredictability and company death by economic recession.

The sweet spot to apply point 1 of this advice is somewhere in the middle.

@lobrien Ours are leaning that way too. I'm assuming multiple teams are seeing the challenge "align with a wall of slots" and deciding it wise.

@webframp So not only do I disagree, I think this is one of the great benefits of computerized decision-making.

When a human makes a bone-head call, they can get better but to a certain extent that improvement does not scale. You can teach new people, but they will forever learn their own way and have gaps in their knowledge that we can't know because we can't introspect them.

When a computer makes an erroneous decision, you can pin its innards like a butterfly, understand precisely why it malfunctioned, and rework it. Then copy that change to all instances of it so that mistake *never happens again.*

Unless by "management decision" here they just mean "gut calls." Because computers make management decisions (resource management, allocation, regulation, distribution, action-ordering) all the time in the modern era, to great effect.

@SwiftOnSecurity I sincerely hope we'd at least have the horse-sense to give those channels some names. ;)

@lauren This comes up in Star Trek circles often because fans who got onboard at Voyager, the 2009 movies, or Discovery go back and try to watch the original series and have visceral reactions to the misogyny baked into it.

And my response is... Good. That's a good response. The fact the show was considered "progressive" by the era's standards and makes you feel that way means it did its job. Because we knew society back then had a long ways to go and we know society now has a long ways to go.

My greatest hope for the next generations is they look on our entertainment with some confusion and revulsion. If we're doing our job right, the past should be a Goddamn foreign country for the people of tomorrow's sensibilities.

... my prediction is that we'll probably see some nodes converge around a common language / expectations of utility, and then outliers.

"Oh I like this one... it offers the Base Kit but also the Search Extension. Don't know how I feel about the AutoBan shared-list banning though."

An issue that will be both a positive and negative for #mastodon moving forward:

Mastodon is a protocol, not a single service. This means that what individual *nodes* do on that protocol differs based on the node you're on. Mine, for example, allows full-text comment searches and following topics by regex, not just hashtag.

This will be a pro because the ability to build different features atop the protocol will encourage development and diversity. Never care about spoiler tags? This node lets you disable them. This node lets you regex for which ones you auto-show and which ones stay hidden. This node lets you specify trusted users where spoiler-tags are auto-shown, etc. Let a thousand flowers bloom.

This will be a con because users actually hate complexity and it takes a particular mindset to be excited about the answer to a simple question of "Does Mastodon support X" being "It depends."

@lauren In her defense, it's *hella*-racist.

("It was a different *tiiiiiiime!*, says the overlay-apologetic Disney fan in me ;) )