Notices by cliffle@hachyderm.io

@dalias sure, you don't _need_ it, but I do think it's a particularly exciting development for non-cartesian systems like this!

@dalias ah, yeah, I've never gotten those tuner apps to work properly (but I have a stroboscope because nerd).

Well, neat! I'd love a proper delta bot, particularly with the recent advances in non-planar slicing algorithms (since delta bots don't move in planes). I will be following this closely.

@dalias looks like the axes are belt-driven -- I've found belt tensioning to be a difficult process for folks who haven't done it a lot, and I'm curious if you've got thoughts on it.

@dalias I don't think the defaults are especially server-specific.

I used to use the defaults on machines with spinning rust disks that were mostly idle -- since each time you spin up / spin down the disk, you're effectively shortening its lifetime.

The defaults can also make sense on a machine with battery backup where disk writes are power-intensive. I actually wound up lengthening them on some laptops, to coalesce small writes into larger IOPs. Not sure if this matters anymore with good SSDs though.

For removable devices specifically, I've seen people use udev to hasten writeback _just for them_ without affecting performance of internal disks, but I haven't personally tried that. (I don't find the flush option to make writes unacceptably slow on removable disks, personally.)

@dalias I mean...

"This watch is timeless!"
"...um...what does it indicate then"
"Cosmic muon flux."
"I'll take two please"

@dalias neat tip, thanks

@dalias neat, so you're basically making concrete using the baking soda as the aggregate? Or is it playing a chemical role?

@dalias this would go great with my registry of mid-grade dependencies that you can use, I guess, but no one exactly loves: sosocrates

@dalias fortunately, there are two kinds of TLS clients in the wild:

- Those that accept gzip and brotli content encoding
- Those that are handrolled and probably trying to evade fingerprinting.

You can just block the second category. It's like wearing a sign saying "I'm a bot please stop spacebombing me."

@dalias I might have some encoding based defenses on some servers. Brotli is *great* at this. Hypothetically.

@dalias there are a couple other lineages outside the hobbyist space (like Stratasys'), but otherwise 💯.

Orca and PrusaSlicer cross-pollinate often enough that we could almost collapse them to one, I suppose.

@dalias Unicode: it's complex, it's imperfect, it's frustrating, and it's _so much better_ than anything that came before. 💯

@dalias I'm really looking forward to seeing what you come up with here.

@dalias also, under the right circumstances, new functions that _are not overloads_ but alter the behavior of argument dependent name lookup ("Koenig lookup")

@dalias fair. Sometimes it can be difficult to predict which interfaces wind up being external, of course. For instance, if one component can be tricked into skipping a validation layer and handing a block of code to a more internal component, any issues in that internal component could be accessible.

Really though, what we're getting at is that distinguishing internal from external can be hard and takes a judgement call.

@dalias While I think you've got a point here, the main caveat I'd offer is around security vulnerabilities. I keep finding "no dependency" programs (which in practice means they vendored, or just copy-pastad, some code) with out-of-date and vulnerable versions of things.

If every program did this, the process of stamping out those vulns would be very involved.

@dalias I mean, it _is_ a vulnerability, just not in _your_ threat model. (Or mine.) They should probably not have included arbitrary memory read/write commands in the HCI. Seems sketchy.

But this isn't going to allow drive-by rooting of your coffee pot or whatever, as far as I can tell.

@dalias the more detailed discussions use the term "HCI," which would be the _host controller_ interface -- the serial link from the ESP32 to the device it's installed in. This suggests that it's a way to root the (certified, fixed-firmware) bluetooth module from a device it's installed in, which does sound useful, but is not at all a remote-accessible backdoor.

All the advisories are damn short on details though. I could be completely wrong.

@dalias hard agree.

@dalias you doing a 4S pack?