knack

@_thegeoff@mastodon.social @matildalove@wetdry.world @rgarner@mastodon.social I archived the top 50(?) or so on limerickdb before it was lost. I think my fave is A woman in liquor production Owns a still of exquisite construction The alcohol boils Through magnetic coils She says that it’s “proof by induction”

@nonehitwonder@tenforward.social "Rewards for psychopathic behavior" are an emergent property of a system. Given a system, almost all attempts at predicting which behaviors will end up being rewarded by them get it wrong. We need ways to make those changes _continuously_ - but anyone who desires power over others will immediately spot that they could join the institution which makes those changes.

@nonehitwonder@tenforward.social it’s proven enduring hard to stop that from happening. I have come to believe it’s the single hardest problem in politics. Organisations formed to get rid of them frequently end up being run by a different psychopath who ends up rich. How do we actually stop it from recurring?

@jameshowell@fediscience.org The increasing exasperation matches "What have the Romans ever done for us".

@hacks4pancakes ooh, I want it. Pun about the term "DM" being unclear in this context elided for brevity.

@sabik @futzle In jurisdictions I'm familiar with, 'actively destroying evidence' is treated very differently from 'failing to act to preserve evidence'. I think whether that works would hinge on how tech-ignorant the judge is / is not - not something I'd stake my personal freedom on lightly.

@futzle so, a “hostile admin” (or “coerced by threats”) is, I think, not possible to forwards-defend against in a federated system(that is, new follows made after they turn malicious cannot be protected from snooping). They could (eg) store any e2ee keys, etc. E2ee breaks search and the local timeline (some homomorphic encryption methods might make authenticated search possible, but it won’t be cheap or easy). Encryption at rest with a regular proof-of-life required to avoid scrambling the keys could protect from seizure by incompetent cops, but then you randomly lose your instance if the admin goes away briefly (and competent cops will physically freeze the ram before disconnecting power, so it doesn’t lose state). You would need some way of detecting movement near the server and requiring reality every time it happened to offer any protection from that.

@futzle in a secure-vs-seizure design, failing to hit the switch in time would delete the key irrevocably; you could not produce it later, it’s gone (along with all instance data). You _could_ use partitioned storage with postgres to put only non-public posts on the encrypted storage.

@chrisamaphone @tinybird @iris_meredith one contrarian data point: my proprioception has always been near-perfect (eg I can use my hands effectively while blindfolded), but my interioception is almost nonexistent (have to focus to tell whether I am feeling bad, very hard to figure out what kind of bad / what part of me hurts). Perhaps “helpful, but not the whole picture”?

@dymaxion I think the group needs at least one 'tactically minded thinker', but beyond that nothing struck me as "you need to think in security terms in order to figure this out"

@hacks4pancakes

Great introduction section, looks pretty easy to port to another setting or system (handy if you have an established game to add it to).

The meaning of attuned to bearer's heartbeat (silver keycard) is unclear to me.

I'd definitely raise the alarm much faster after a guard is found knocked out - especially after a potion is administered to wake them up, and they can describe their assailants. IMO that's "page the head of security, ask the watch to be nearby, double-check that your surveillance is working, begin a sweep of the building" territory.

On layout & indexing:

There's a few tables of "thing players might try -> information they can get on a success".

I find the inverse table extremely useful when running a module - a list of all the useful facts/accesses a player could obtain, and all the designed ways for them to get it. Especially useful when figuring out how to nudge a group who have gotten stuck.

Very similar to the Counter column in the "Magical Security Devices".

GM can cross off problems the party has solved - so they can clearly see which ones they haven't solved.

@tante “collected writings of the internet”, not “collective knowledge of humanity”

@davidcelis@xoxo.zone just a quip

@debcha@mastodon.social Is there much difference between bookstores RE how much the author ends up seeing?

@xssfox@cloudisland.nz If those spikes in s3 costs are from folk syncing the entire s3 bucket, getting everyone to connect to an endpoint outside AWS instead (and have that cache your s3 bucket) is the only way you'll bring those down while keeping the data public; AWS do not, as a rule, offer discounted egress other than lightsail.

@xssfox@cloudisland.nz I've spent quite a bit of time on cloud-cost-stuff. If keeping the architecture you have with AWS, you could _maybe_ squeeze out a 20-30% saving (downsizing any overprovisioned stuff, savings plans / spend commitments), but it'll keep growing with usage. You could enable 'user pays for requests' on s3, but then the bucket is only accessible to paying AWS customers.

@xssfox@cloudisland.nz Opensearch pricing is _wild_ - AWS charges almost double their on-demand instance pricing (which is already egregiously high) merely to install opensearch in a managed configuration. EG the price you're paying would get an m8g.xlarge.search from AWS (4 cpu threads, 16gb ram, slow storage at additional cost via EBS). An AX41 from hetzner costs 3 times less, and has 12 cpu threads, 64gb ram and comes with NVME storage (which is hundreds of times faster). Looks like you're pretty locked in to AWS though, with sqs / sns / lambda / api gateway.

@xssfox@cloudisland.nz I don’t know your architecture, but you can definitely use SQS / SNS queues from outside aws

@xssfox@cloudisland.nz I had a read of the project docs / terraform. If I'm reading those correctly, it looks like data comes in (reasonably fast) to https://api.v2.sondehub.org (api gateway -> lambda -> sns -> sqs -> elk (opensearch?)). Is all of that infra there to avoid losing writes when incoming data is arriving too fast for ELK to keep up? I also see a bunch of SNS -> mqtt stuff in there, so maybe it's also for fanout to other services?

@xssfox@cloudisland.nz I can't speak to the fanout part, haven't made sense of what else it's fanning out to. How much faster would writing to opensearch need to before you didn't need a queue between incoming requests & the database?

@xssfox@cloudisland.nz yeah, I get that. It's a lot of faff to set up a single instance, and being able to do failover doubles the cost and triples the faff. All of that said, even with three servers (one in service, one hot standby, and one offline for upgrades) I'd expect you to halve your current spend _and_ have headroom to scale up by a factor of at least 20.

@xssfox@cloudisland.nz yeah. AWS are great and I use them a lot, but they charge an absolute fortune compared with any of the dedicated server providers. Hetzner Robot are the cheapest $/performance out there, but they only have two DCs and you might want more geographic redundancy than that.

@xssfox@cloudisland.nz I know you've got a lot of sensors reporting in, but... how many gbps is that peaking at?

@xssfox@cloudisland.nz Not trying to talk down to you; I'm familiar with at least some of your work. Just getting seriously nerd-sniped by thinking through what's possible.

@xssfox@cloudisland.nz cheaper to use DNS load balancing across nodes with 1gbps links than a 10g link (but I really need to stop thinking about this and go do stuff that's _way_ less interesting)

@davidcelis@xoxo.zone “this conversation is limited to collaborators” has never been a more apt bit of UI copy