HNHacker News
TopNewBestAskShowJobs

archi42

1,719 karma · joined November 25, 2017

he/his. Computer Scientist working in cyber stuff
submissionscomments
archi42··on Ask HN: Robot vacuum but without the privacy invasion?
Uh, I think my friend ordered one locally and (because that was the last one in stock) another one for his parents from AliExpress.

I didn't do the replacement, but it seems reasonably simple - the cylindrical sensor unit has a cable with a 4 pin connector and is clipped in place.

The new one doesn't look exactly like the old one and had different makings. I hope this means it now doesn't randomly fail.

archi42··on Kenton's House
It's not that crazy ;-) I actually emulated your approach for release testing. Except I was using qemu VMs.
archi42··on Kenton's House
Yeah, I remember the copying. At some parties we spent a lot of time doing that.

Some CDNs use HTTP with post-download verification (I hope). I suppose local caching is the reason for that decision; think big LAN parties with a "only" a few gigabit/s of connectivity (if at all).

A nice solution is what Steam does: The client checks a magic DNS entry (lancache.steamcontent.com), which you're intended to override [0]. It then just works.

For Riot, Blizzard, Epic and Windows Update I've set 36 overrides in dnsmasq (they supply a script for that); those are branded akamai hosts and such. It supports more launchers and platforms/consoles, but from the big list [1] that's all we needed.

Now GOG uses HTTPS, so for the reason you stated that can't really be cached for random 3rd party machines. Subsequently, if LanCache gets an HTTPS connection, it simply proxies the encrypted connection to the correct server using SNI sniffing.

Regarding bandwidth: Two games on our list were CS:GO and LoL. IIRC four people had to install CS:GO and six people had to install LoL. We could have setup file shares, copied the data (as usual, with multiple people saturating the first share to come online, and nobody using the second or third share), hope laymen know where their steam library lives and to put the correct folder there... or just let them download the games directly from the client, except on a virtual 10 Gb/s pipe ;-)

Though to be fair: In Steam you can configure the client to allow others to download their games from you on the local network. If enabled, you can configure to share with "only me", "steam friends" and "everyone". They added the feature earlier this year[2]; probably for the Steam Deck? One of the attendees installed Baldurs Gate 3 (>100GB) from a friend that way. However, the local download feature wins over the lancache, which messed things up a bit when everyone installed CS:GO and pulled that from a single PC instead the server; so we had everyone disable that.

I'm now running a smaller cache on my home server. Additionally, I have a spare server sitting here that's collecting dust. I think I'll permanently designate that my LanCache server for future LAN parties.

[0] https://lancache.net/news/2020/01/14/steam-client-now-suppor... [1] https://github.com/uklans/cache-domains [2] https://help.steampowered.com/en/faqs/view/46BD-6BA8-B012-CE...

archi42··on LAN Party game caching made easy
The recent discussion of Kentons LAN party house[0] reminded me to share this - since for some reason it seems this was never posted?

This acts a simple to setup HTTP cache for several gaming services like Steam. Just start two docker containers (one cache, one DNS), point the DNS server to the cache and its set. For better preparation, there are some tools to pre-warm the cache.

Thanks to affordable switches with SFP+ ports and a pre-warmed cache, we had several people simultaneously install games at over 100MByte/s - on a 100MBit/s connection.

https://news.ycombinator.com/item?id=37366584

archi42··on Tidal energy is not renewable
One thing I notice is the assumption of unlimited exponential growth by 2% per year. That's a huge fallacy. Quick check, yeah, 1.02^1031 = 735,829,316. I'm pretty sure we will be using magnitudes more energy than today because we're more people and hopefully with better living standards for everyone. But even then that's a lot. And on that scale I'm not entirely sure where the whole energy should go to... Maybe produce mass/objects out of it?

If we reach peaked our energy consumption in merely 250 years, that's less than 150 times our current consumption. I didn't do the math, but would date to suggest this gives us a few years more time on this planet.

archi42··on Kenton's House
Updates? Installing games? For our last LAN I brought a machine with LanCache[0] installed and mildly warmed up. I convinced the host to finally unpack his new switch with SFP+, connected a DAC and had the host point the DNS to that machine.

Having a bunch of people install cached games at GbE speed all at once, on a 100MBit/s internet line, was pretty nice :)

Obviously with 10GbE this is limited by disk speed. But a minimal Linux can buffer a few games in 64GB of main memory.

I had it cache Steam, Epic, Riot, Blizzard and Windows Update. The cache was warmed at home over night, based on our list of "games we might play". Though Steam was limited to games I owned.

[0] https://lancache.net/

(But yeah, I totally get what you did and would do the same!)

archi42··on Ask HN: Robot vacuum but without the privacy invasion?
As much as I like the L10su (recommended it to a friend and put Valtudo on it for him), caveat emptor: The ultrasonic sensor used for carpet detection is prone to fail. Most detect no carpet anymore (and proceed to mop it), others detect it everywhere (and don't mop anything anymore).

You should be prepared to sending it in for a warranty repair, probability for failure seems to be 5 to 10% from what I notice on the Valetudo telegram channels. Check Google for other peoples experience with Dreame repair centers. He replaced the sensor himself, but that breaks the warranty seal.

Others than that, he's absolutely happy with it.

Check https://valetudo.cloud/pages/general/buying-supported-robots... for the list of recommendations. I've a L10pro myself and it works well, too. But looking at the prices, I'd get a D10s plus or pro myself these days (that is, if the L10su wasn't an option, since it has the superior feature set; I'd be willing to take the risk with the sensor and just swap it).

archi42··on Kate Editor Features
First hit using Google: https://www.reddit.com/r/golang/comments/cokcuw/the_kate_edi... LSP support for Go was introduced 4y ago. There were some discussions regarding go-langserver vs gopls, but that could be dated (Christoph & contributors are quite active).

Personally, I'm an e-vi-l person and never worked on Go projects, so I can't tell you from experience. Also I don't know your requirements. Maybe just give it a shot? After all, its FOSS, so the only investment is your time ;)

archi42··on Kate Editor Features
Yes and beyond: You can also use it for full software development. After all that's what it's intended to do. It's no accident 6 out of 8 screenshots show source code. I know/knew some pretty skilled devs using it for their defacto-IDE ;-)
archi42··on Heat your house with a mechanical windmill (2019)
My key takeaway: GP is right, going PV+HP is a safe default; at least around here(!!). Unless you don't get money for selling excess power to the grid (batteries have become massively cheaper though) AND you have a gigantic roof. Even then, it's probably worthwhile investing more into the PV instead of getting solar thermal. Of course this also depends on latitude.

Anecdata 1: Thermal solar installation between 49 to 50deg latitutide, central europe. My uncle's SFH.

Anecdata 2: We're currently adding PV+HP with cooling capacity to our own SFH and I would have loved thermal solar to be great, so I looked into this. (same rough geographic location).

During summer his thermal solar has great efficiency, but the heat tanks are already charged after 1 to 2h (that's what he said when we talked about it last summer). So the installation has at most 2h of say 80% efficiency (during low solar radiation in the morning!) and then another 8h of 0% since the heat can go nowhere. This also wastes the huge energy peak after noon. This means: The ratio between a days "total energy harvested" vs "total energy which reached the collectors" is even less than the arithmetic average of 16% efficiency. Of course they use hot water over day, but that only increases the efficiency a little bit. Especially considering the physically huge tanks they have installed (probably 1000l ballpark, and I think there are two of these?!). Contrary in winter, the temperature differential between collector and fluid is so low that no heat/energy can be harvested at all. (At least for their installation and the winter temperatures around here).

Compare that to PV+HP: Any solar radiation is transformed to electricity with 22% efficiency. In summer the COP [see sibling comments] is very high, so you can still quickly charge that hot water tank and then use the remaining energy harvested for your home or car, or sell it to the grid. Plus the hot water tank can be smaller (less losses!) since you don't need to "store" all the excess energy. E.g. we have 200l hot water and 200l buffer for heating (or actually: Defrosting the HP). During winter the PV still produces some power that the HP can convert to heat (not too much and especially for less hours per day, but >0).

Maybe there are some sweet spots during spring and fall when solar thermal outperforms the PV+HP because it takes the whole day to charge the tanks. But then again when I installed my PV this spring, it produced >110% on a cool (16degC/60degF?), sunny day. So even during those seasons PV+HP are a tough competition.

Nice bonus for the PV+HP (that's what we're doing): If it's too hot in the summer, you can use the excess electricity to drive the HP in reverse mode for 7degC/44degF water temp in the "heating" system for active cooling. Due to condensation this requires dampproof insulation for the pipes, and convector units (sometimes also called "fan coils"?) to move enough air through the heat exchanger (car-like radiator) instead of traditional radiators or floor heating. But every convector can sink 1500W (5100 BTU?); for a room in a well-insulated house this can be enough.

Mind that our roof (south east + north west) has merely 60sqm per half, minus windows and other installations, so there is not too much space to add thermal solar. We get 0.07€ per kWh sold to the grid, so the excess PV mostly pays for itself. Battery is 11kWh usable, based on budget not "what we need" (data wasn't sufficient because we did all the renovations at once); this improves how much we can consume ourself; was more worthwhile than adding hot water tanks (we just can't consume that much hot water every day). We're also adding massive insulation to the 70y old house to slash heating requirements from >40MWh/y to something more sensible.

archi42··on Jazz² Resurrection: Open-source Jazz Jackrabbit 2 reimplementation
Thanks for digging this detail out! I'm a bit disappointed since I was looking forward to playing this with friends again (fond memories), but I can totally understand how the dev might not have the necessary time to implement this.
archi42··on Developers are building communities that act as their own miniature power grids
Natural gas stoves are unhealthy[1]. Poor people already have worse health than rich people for a multitude of reasons [2]. So if more poor people get access to places with electrical stoves, their health improves.

I only looked at walmart.com: Cheapest gas stove was 122US$, cheapest induction 222US$. If the 100US$ difference is relevant, you're neither building a house nor looking to rent an apt without a kitchen anyway.

[1] https://www.google.com/search?q=natural+gas+stove+health [2] https://www.google.com/search?q=health+of+poor+people e.g. https://health.gov/healthypeople/priority-areas/social-deter...

archi42··on Alcohol without the hangover – scientists are finding ways
The Bitburger NA with the blue branding is good. Don't be fooled if you don't like the normal Bitburger, imho their NA is much better :P OTOH, if you happen to like Bitburger, they also have a newer "dry NA" variant with green branding, that tastes much more like their alcoholic beer. No idea about international availability, but at least a slight chance.

For wheat beer, the bigger German what beer breweries offer a NA variant these days, and those are quite good. I'm thinking of Erdinger and Paulaner, which are probably your best bet internationally. Don't tell anyone I know, but my secret recommendation however is Öttinger NA wheat beer. Those are cheap and taste very good (fellow Germans: If you don't believe me, do a blind tasting). But don't ever the mistake of buying or even drinking their normal Pilsner/Export beer - those taste like pi... ;-)

If you're ever in Germany, is "Mönchshof Radler Alkoholfrei". Radler is "shandy", a mixture of beer and lemonade. I've yet to meet anyone who likes beer but not this.

archi42··on Alcohol without the hangover – scientists are finding ways
This is also what surprises me: No hangover and no liver damage are great, but doesn't say anything about physiological or psychological addiction. Especially since the lack of any short-term negative feedback loop (alcohol poisoning, vomiting, or at least a hangover) makes forming problematic consumption patterns much more likely.

OTOH, "one a day for a week" is very unspecific: Is that the equivalent of drinking a six-pack or two every day, alone; or gathering with friends to taste a literal sip (e.g. 5ml) of spirit? Do they act equally (not only in strength but also in decay and other factors - maybe the medicament has a booster or some chemical delay). Is there a selection bias among people who (self-)medicate, do they have traits or common environmental factors that make negative outcomes more likely?

I don't want to overly defend this, since in my usual peer-group bringing that stuff to a gathering would feel a lot like "let's do drugs together!" (only few gatherings involve drinking any amount of alcohol) and I personally don't want to figure out empirically that instead of a hangover this triggers even a light depression (or worse) - but the comparison still seems a bit simplistic to me.

archi42··on Bitsummit 2023: Steam Decks everywhere, no VR, and impressive indie games
I liked the 2017 Rick and Morty game. It's rather short, iirc about two hours.

We often used to hang out at my old place with a few friends (often the first days of the year, but also some weekends), playing boardgames, cooking together and taking it slow. We played the game on two or three of these occasions and having one person do a full play through with the others watching/commenting on the TV worked incredibly well. Plus, spectators don't need to focus on the game. I think one time we did two play throughs back to back (different players, and I think the second one wasn't there for the first round). Though the living room at that place was huge and allowed for a 3x4m arena with only minor inconveniences. And of course if people loath R&M it's less funny I'm afraid (but still a very nice VR puzzler!).

archi42··on Study finds billions of nanoplastics released when microwaving containers
> [...] all plastics impart a taste to water once the water has been sitting for some time. I imagine heat speeds up this process. [...]

I feel the same. Water from most plastic bottles tastes very different from the same brand & "type" in a glass bottle. Thinking about this, I should eventually do some experiments (e.g. ABX and fully blind). Especially since summer heat should impact this during transport.

For plastic dishes, I avoid these for hot(!) foods as well. At least my SO's cheap plastic outdoor dishes add some flavour. Though for fridge storage or cold foods I never noticed the plastic flavour. I still prefer glassware for practical reasons (except weight) and (highly subjective) fell like it's just nicer, but e.g. at BBQs I (luckily!) nothing ever tasted like plastic only because a salad was in a plastic bowl; same for use-once plastic stuff used at some events, luckily the short contact never contaminated the taste (which means this could be placebo except for really bad plastic ware!).

Just out of curiosity: Do you also taste milk changing its taste long before others notice? Maybe we have a similar sensitivity that's also affecting "plastic flavour"?

As for hot spots in the microwave, that's just a drawback of the tech imho (or our old-ish, non-inverter microwave with a poor radiation pattern), and except for reducing these (70% and stirring), I live with that. Though we basically never prepare raw food in the microwave, only reheat leftovers or boil water/milk. So actual food safety is less an issue for us. On especially lazy days (and with highly unevenly heating food like yesterdays home-made lasagna) I even skip the cold parts and put them in the microwave again; I'm pretty sure this is safe.

archi42··on Hunting for Nginx alias traversals in the wild
If that's an option then that's the right way to go. There is a reason some MTAs have been doing something like this for decades now (I'm thinking of qmail).

To be honest, I'm not sure if it's even possible to run the application/interpreter/cgi (e.g. php) as a child of the nginx process - though with Apache I'm still seeing that occasionally.

archi42··on Hunting for Nginx alias traversals in the wild
Most (I hope all) distributions already run nginx as a separate user. It's best practice.

But that won't help if you alias to "/foo/bar/www" and the the application has a SQLite database at "/foo/bar/db.db", which the nginx user has to have access to. Same if you run it in a container (or lock down permissions using systemd).

archi42··on Btreefs generates executable code at runtime to unpack btree nodes
Ah, yes, that's an excellent point. Because someone else talked about 5% performance I looked at it with my performance hat on, not the security hat. OTOH we have BPFilter with its VM.
archi42··on Btreefs generates executable code at runtime to unpack btree nodes
The general optimization is actually not that novel: A DBMS might do that for parts of a query. At least in a high performance database lecture this was taught as a possible optimization. Edit: I'd intuitively expect the improvements to be more than 5%< though.

E.g. PostgreSQL can do that these days (not sure if it did back then): https://www.postgresql.org/docs/current/jit-reason.html

archi42··on “Exit traps” can make your Bash scripts more robust and reliable (2013)
Thanks for pointing this quirk/difference out! I'll keep it in mind.

Regarding the expected behaviour: I believe this is largely a matter of preference/philosophy. I have previously used trap handlers in bash scripts explicitly so that the cleanup is run in every case and /tmp is not polluted - even of failures. Reasoning: For automated tasks or people not familiar with Linux/my script, the disk should not be filled up with old garbage data. And for debugging, I usually add a '-d/--debug' option to have more verbose output and disable the cleanup on Ctrl+C (but still cleanup on normal exit).

But as I said, I don't believe there is the one true way. So except for pointing out that I had to ensure external cleanup in case the script was run by some auomated tasks in a permanent environment (read: native), I wouldn't seriously complain if I ever used one of your scripts :)

archi42··on Missing Titanic sub faced lawsuit over depths it could safely travel to
The forces are immense and act all the time. In your daily life, you don't usually have that kind of system, so this feels counterintuitive.

Press against a solid object gently with one hand on each side, then suddenly yank away one hand. The object will be accelerated gently. Now do that with all your strength and send the object flying through the room in no time, with a few dozen kg of pressure. Now imagine that you could apply hundreds of tons of pressure. In such a system, every change will be nigh instantaneous.

archi42··on Ken Thompson on the bug that exposed his compiler trojan (2021)
Because it was a bug. Maybe he always appended a trailing zero byte ('\0') to some string.

In general, if you're not aware of the content of the paper, you might want to familiarize yourself with it. In short & maybe simplified: The pre-compiled compiler contains a malware that, when compiling the compiler source, will be insert itself again. (Of course the source does not contain the malware). They used the pre-compiled compiler ("stolen" from Ken) to compile the compiler's source code; this of course inserted the malware into the newly built compiler. Then they used the newly built compiler to compile the compiler's source code again, and so on. And every time they did that, it grew by one byte because of a bug in the malware component.

archi42··on A profit-price spiral in Germany
Seems a bit hugged right now...

http://web.archive.org/web/20230612200920/https://think.ing....

archi42··on Nvidia releases new AI chip with 480GB CPU RAM, 96GB GPU RAM
They can not be made "anywhere"; when you can't get that PMIC from the original manufacturer, good luck getting it from someone else. And replacing an IC in a QA tested, EMV verified, FCC and CE etc. certified device will often trigger you redoing all that, possibly requiring additional iterations. If there is a similar part available at all.

Take a look at a recent GPU and count the auxiliary components. All of them can cause supply chain difficulties.

archi42··on Electric cars prove we need to rethink brake lights
Some do: In ACC mode my 2013 Volvo uses radar to "lock" on the car in front of me. If it is slower than the ACC set point, its speed is indicated in the instrument panel.

It looks like this: https://www.motoreport.de/wp-content/uploads/2014/01/IMG_296... Set point is 110 km/h (green marker), the car in front is going 80 km/h (solid white line indicates the speed difference). The small red triangle indicates the speed limit of 80 km/h. Obviously the ACC is also cruising at 80 km/h, so most indicators are focused there.

And you're right, even with the limited form (single car in front of me on the same lane, data only shown if slower) this is amazing.

archi42··on Tell HN: Be aware of people trying to scam contractors
`sudo -k` is your friend ;-)

But for an attacker that convinced you to run some malicious shell commands (e.g. `curl ...| sh` or copy/pasting more than you bargained for) that's only a minor inconvenience; next time you sudo, they're root anyway.

//Edit

Also, an attacker aiming for root is a red herring. One compromised account can already be bad enough - why would they need root on your private computer when they already stole your password database, keylogged your master password and created an off-site backup of your home dir?

archi42··on Ask HN: What is the best password manager available today?
I was in the same boat as op: Didn't want to care about sync at all and use it on all my devices. Didn't want to rely on a third party. Vaultwarden solved that for me.

Like all services I self-host for personal use, it's only accessible via VPN.

archi42··on Only one pair of distinct positive integers satisfy the equation m^n = n^m
They say just before that: "Then we can say that 1 < n < e."

Since e = 2.71..., there are not too many options for n. The sentence is a bit overconvoluted, but it's correct.

Of course if the search space is bigger, you need to either do some more thinking to narrow it down - or spend a lot of time probing candidates.

archi42··on I made my blog solar-powered, then things escalated
I have it running in a testing setup for about 2 weeks now. 8 modules flat on the lawn (Maysun glass/glass 410Wp), the battery connected and tied to the grid with export power limited to 0 because my meter is ancient enough to spin in reverse (which is illegal to do here).

Which setpoints do you mean? I'm not familiar with all the terms, especially when they vary by manufacturer.

If you mean the power limit the Deye should put on the DC bar (i.e. put into the battery), that works reasonably well. If it's at 100% it will still charge a few Watts, but nothing severe.

The other meaning I could come up is the target power to draw from the grid when export is disabled (as in my case, for now), or when it is feeding energy from the battery into the house when I draw more than the PV array delivers. That's 20W by default, but can be user configured. This fluctuates somewhat strongly, but overall works out correct. I believe the reason isn't the software, but the poorly routed measurement cables: It uses measuring transfomers and the cables in my setup go tightly and parallel to the power line that connects the inverter to the grid. The electrician said that's not an issue, but I guess he usually measures 50A+, in that case the noise is negligble. Around the 0A point the noise has a much bigger impact. But that's easily fixed. Eventually I'll use that option to prevent the battery from discharging when the car is charging over night (e.g. when I charge with 1200W, I plan to set the target to that value).

I'm not sure if I can configure other setpoints, e.g. the MPPT target voltage (which would be a nice trick to get shadow managment). I can define some other things like "from $x a.m. to $y p.m., keep the battery at $z%, grid charging is [not] allowed, powering on the external generator is [not] allowed". For testing the battery balancing (it has an additional Neey active balancer) I used that. And yeah, it can trigger an external 3phase generator to power up (not sure if the 1ph can do that).

As far as I could find, all options are exposed via RS485. I adjusted the sunsynk HA addon to work with the newer 3 phase inverters (sunsynk = deye rebrand; changes are already merged upstream). Only thing that I'm missing is the current limit set by the BMS, as well as other BMS data sent to the inverter (I did not search the published register map for that, yet).

What I can already say:

All in all the technical side is pretty nice; as said in a sibling comment, I could theoretically push out 12kW AC on a single phase if my local load is severly inbalanced. And the MPPTs are pretty flexible in how much power they can accept (as long as the sum is less than 15.6kW at all times).

What's horrible is the UI side. Some phrases are highly unclear, and requires knowledge of what the option actually does. E.g. "signal island mode" means "in case of grid failure, tie neutral to PE on the UPS output". In German installations that's always necessary, else RCDs will be much less effective. They would benefit greatly if they had a team of knowledgable translators.

Having to ask their support to do an update is also a joke; especially since they don't publish release notes. Also, the update power cycles the inverter, which causes the UPS output to lose power.

My 14.3kWh LFP battery has a 200A Seplos BMS. The inverter and the battery talk via CAN. In the BMS the power limit is set to x Amps. But the inverter shows a limit of (x/2) Amps. Now my battery should eventually operate a 140A continuous, but I'm not feeling to well about setting the limit to 280A and one sunny day having the two systems decide to get along and actually push 240A (the Deye's limit) over the connection+BMS sized for "only" 200A. The solution is to get another BMS, but back when we formed the respective group orders, people didn't yet know of this incompatibility. Plus we were focused on Seplos because the Victron users felt it was imporant to get a BMS that would not disconnect the negative bus.

Still, for someone looking at a small plant with just a few panels and a small battery, it's a great value preposition. Especially when paired with a different battery/BMS. With the 8 modules behind the house (shadowed in the morning and the late evening) I had 105kWh PV and 120kWh grid power over the last two weeks. That's 1k€ for the modules, 1.3k€ for the smaller 1phase 3.6kW inverter, 1.2k€ for the 3kWh Pylontech US3000C, plus some cables and AC fuses for 200€, plus a DC fuse+SPD box (<100€ if DIY) = 3.8k€ for basic stuff. What's missing is panel mounting (800€?) and, depending on local laws, an electrician to sign off on everything.

← PreviousPage 6 of 34Next →