I unexpectedly found myself working for the UK subsidiary of AJ just before the .com bubble pop. Interesting times. Things I remember:
I wrote something to do cluster analysis of the previous day’s search queries. It turned out that the most frequent search was something like “naked picture of $soapOperaShowActor”. Actual search query data might shake your ideas of the goodness of people.
Much of AJ’s content was based on editorial staff (often young journalistic folk) researching what they thought might be the highest quality answer. One day I passed the desk of a colleague who was watching porn. What now? It turns out that they wanted to be able to answer the question “best porn of $kink” for a large variety of kinks. Which meant that they also had to have a policy of how to direct queries for CP. To something less harmful obvs.
As a corollary of the above, the editors needed a way to search for candidate results. What did they use for this? Google of course!
Via an acquisition I worked for AJ in the US for about a year before the move to the UK. It was a vivid illustration of the way in which dishonesty and backbiting could permeate an org. I knew plenty of fine individuals there, some who kindly taught me hard lessons, but as a company, a culture, it was a cesspit.
Anyway I got laid off in the great wave of 2001, was out of work for a while, did some truly awful work on supermarket planogram s/w and eventually got a gig doing IP routing. Ever since then I’ve been patronising grad hires by telling them how useful it is to have a bad job in your past. It makes it much easier to cope with occasional bad days at an otherwise good place. “Sure, my code crashes on a double exception when the reverse bcopy chokes on an unwired chunk of address space in the ARP lookup interrupt path, but at least I’m not trying to optimise the positioning of cornflakes to take advantage is this month’s promo pricing”. Good god, there was a time when I had a subscription to The Grocer magazine. Watch out kids. This could happen to you! (I also got to spend a day following a guy around the London Underground as he refilled chocolate vending machines. But I won’t talk more about that unless you buy me a beer).
It's been a while but I also worked there at the same time. I was in the original group who set up the UK operations around the turn of the millennium.
And you recommended the Introduction to Algorithms book to me...
>Remember that 99.9% of people do not consider themselves to be the bad guy, yet more than 0.01% of people are bad guys. Almost no one identifies with evil, yet evil is a string that runs through every beating heart.
"estimate the prevalence rate of psychopathy in the general adult population at 4.5%." [0]
You do most of humanity a disservice by lumping them in with that cohort that may or may not identify themselves as evil (I have no idea) but are certainly capable of deliberately and with calculation behaving in ways that most of us would label with the "E" word.
I think I understand why this is true for plain IP forwarding. There isn’t much to break the cache and the lookups are few and fast.
What’s the cheapest (new) computer that can drive a 1Gb port with NAT? With a busy encrypted (wireguard?) connection?
[I don’t think qos has a lot of use in the domestic environment; sure, someone here does it but I think it’s much less mainstream than the features I already mentioned. ]
Such a device could drive my home. But in a couple of years I suspect I’ll want 2Gb or 10.
In the past I’ve tended to use a device until its crappy power supply failed. So I guess I’m hoping for a >5 year life span/upgrade capacity.
For all I know the answer to my question is one of those passively cooled four port n100 bricks from AliExpress. Anecdata happily accepted.
Wireguard adds nothing unless you'd want to terminate it on the router. In which case it adds so very little it's unnoticeable.
About any n100 will do. Question is in their reliability which mostly comes down to power regulation components quality. Not performance.
One of my installs runs on a repurposed old android phone. Which has about 100 times CPU capacity of the router I write this through, and that one being cheap tplink shit still terminates wireguard at link speed which is 100Mbps. You don't need fancy gear for routing. And you don't usually need gigabit uplink because speed is limited way upstream.
But if you want "the right gear and damn the price" go get a Microtik. They are very good.
> What’s the cheapest (new) computer that can drive a 1Gb port with NAT?
What's the cheapest new computer you can find? That will work. If you have PPPoE, you need to be a bit more careful; depending on your OS and NICs, it's possible for inbound traffic to only use one core; low power laptop cpu may not have enough throughput from a single cpu, but my information is a little dated.
I did 1G NAT on a dual core haswell [1] for a long time.
I can drive 2+ Gbps vrfs, nats, ipsec, complex firewall rules and several routing tables through an Atom C3558. This is just using stock linux kernel networking. There are other services running on that box too.
Depending on details, it can go higher (e.g. without the ipsec being handled on the atom box, and using the 10G ports built into the chip, offload becomes helpful for TCP and UDP flows).
This is traffic in one 10G port and out the other, in this case.
Multiport flows were not tested since they were out of spec for the use case.
This is not a one off - this is a product I built and has been tested in many deployment scenarios. (I can't provide more details due to employment reasons, and I won't name the employer)
I think you have the wrong end of this stick. See the Greenpeace vessel Rainbow Warrior for an example. There have been several iterations of this ship name since the first was bombed by the French secret service in 1985.
First, the US constitution as it currently stands admits modifications. Amendments are version bumps. My understanding is that they’re harder to come by these days.
Second, the constitution may be written but the interpretation is always changing. In particular, the interpretation of laws around restriction of free speech have lots of history of being interpreted in ways that may or may not be congruent with the intentions of the original authors, who’re dead, so we’ll never know the truth of it. It’s only been 107 years since the US Supreme Court decided that anti-draft speech in time of war COULD BE ILLEGAL. Apparently that was partially overturned in 1969.
Thirdly [naming, caching and out by one bugs!] it is far from clear that a written constitution will lead to a durable republic. It’s only been ~250 years. Too soon to tell.
> Second, the constitution may be written but the interpretation is always changing
It’s okay if the change is because you think the new interpretation is closer to what the constitution originally meant.
It’s democratically illegitimate to change the interpretation otherwise. A written constitution is already an impingement on democracy. But how can it be that whoever is doing the interpreting is allowed to restrict democratically adopted laws in ways the constitution didn’t originally intend to restrict them?
There is no right to vote in the constitution as written and interpreted in the 1700s. There is also no guarantee of freedom of speech. The first amendment was considered a rule that only applied federally.
What's democratically illegitimate is everything you wrote in this thread.
If your state government threw you in jail for what you just wrote that would be perfectly aligned with your "original understanding" interpretation of the U.S constitution.
For safety critical systems, one strategy is to store at least two copies of important data and compare them regularly. If they don't match, you either try to recover somehow or go into a safe state, depending on the context.
Yes, but what's easier depends on layout. "Consensus" makes me think of multiple entire nodes, and in that situation you can have a nice symmetry by making each node store one copy and one small hash.
If you're doing something that's more centralized then one hash might be simpler, but if you're centralized then you should probably use your own error correction codes instead of having multiple copies.
Seems like chronometers would be a case where two are better than one, because the mistakes are analog. If they don't exactly agree, just take the average. You'll have more error than if you were lucky enough to take the better chronometer, but less than if you had taken only the worse one. Minimizing the worst case is probably the best way to stay off the rocks.
You can have voting systems in place, where at least 2 out of 3 different code paths have to produce the same output for it to be accepted. This can be done with multiple systems (by multiple teams/vendors) or more simply with multiple tries of the same path, provided you fully reload the input in between.
I’d upvote this a hundred times (which seems very in keeping with modern democratic (small ‘d’. Put away that axe Eugene) norms) not because I love the solution, which is necessarily janky, and not because I like the author’s prose style, though I do, but because it ought to be a cause of embarrassment to Apple that this sort of folderol is needed.
Art is only interesting if it elicits an emotional response in the viewer. Otherwise it is illustration.
And the wonder of it is that we can all have different responses to the same thing. (The Mona Lisa is a waste of canvas and oil - a hill I will die on).
I cynically believe that many people will force themselves into having an emotional response if the art piece matches with what they understand as having currency with the type of people they seek to emulate and the rarified scene they want to be a part of.
I think I read here on hackernews that the Mona Lisa doesn't look at all like it did when it was freshly made. If I look at the restored copy https://en.wikipedia.org/wiki/Mona_Lisa_(Prado)#, I at least find the silk very nice.
Your quote needs a “sometimes”. For every murderous, blood-soaked dictator who experienced pitchfork-o-clock there are several others who died peacefully in their beds at a rote old age. Louis the n-teenth lost his head. How about Louis 1-(n-1)?
Anyway I got laid off in the great wave of 2001, was out of work for a while, did some truly awful work on supermarket planogram s/w and eventually got a gig doing IP routing. Ever since then I’ve been patronising grad hires by telling them how useful it is to have a bad job in your past. It makes it much easier to cope with occasional bad days at an otherwise good place. “Sure, my code crashes on a double exception when the reverse bcopy chokes on an unwired chunk of address space in the ARP lookup interrupt path, but at least I’m not trying to optimise the positioning of cornflakes to take advantage is this month’s promo pricing”. Good god, there was a time when I had a subscription to The Grocer magazine. Watch out kids. This could happen to you! (I also got to spend a day following a guy around the London Underground as he refilled chocolate vending machines. But I won’t talk more about that unless you buy me a beer).
reply