Posts by Malcolm Weir 786 posts • joined 23 May 2007
I don't think this is true: US companies like Ireland for a couple of reasons aside from the usual tax/workforce/cost issues: first, there is no language barrier, and second it's an hour closer (timezone-wise) than mainland Europe.
(There's also a third, which is customs pre-clearance; don't underestimate the value of having parts of Dublin -- and Shannon -- airports effectively inside the USA...).
"Responsible" is doing a lot of heavy lifting in that last para!
The flashcards are quite obviously from "kids" (people less than, say, 25) who are trying to complete their required training courses. They are, of course, in some way "responsible" for the security of the weapons (in that they're likely to be the ones holding the gun pointed at an intruder), but what this really shows is a management failure....
Just to follow up: I've looked into this a bit more, and what's really going on is that by putting the memory "on" the processor chip, Apple is saving assembly costs for a penalty of having more CPU SKUs. But as "they" make the M1, that's less of a burden that it would be for someone buying an Intel or Samsung CPU: Apple can decide how many M1's they'll make with each memory capacity, and go from there.
So this is a production benefit (because they've eliminated the SO-DIMM PCB, the sockets, etc) not an architectural benefit. I'm sure the fact that it precludes memory upgrades is just gravy...
This is great spin by whoever is making it, but it is suspect from an engineering perspective. What the M1 is doing is pooling the GPU and CPU RAM (all unexciting LP-DDR4 stuff) and pretending it's a feature by suggesting that there's something inefficient by having the GPU with its own specialist memory (e.g. GDDR6) and the CPU with it's memory.
Compare with a low-end GPU like the embedded ones in an Intel processor: these GPUs manage to share the same pool of memory using "just" the DDR4 interface (so no bus to speak of), so what's the difference between this an an M1? Basically, nothing much.
Now, compared to a high-end system with a discrete GPU, having separate pools of memory (each dedicated to the task) usually makes much more sense, as (for example) the video refresh can be optimized in the GDDR in a way that it can't in regular DDR.
So what this boils down to is someone promoting the lack of a discrete GPU as some kind of benefit, rather than the reality of this being a reasonable cost-compromise.
I'm not actually trying to knock the M1; there is no "one size fits all" chip, and the M1 seems a pretty capable part for what it was intended to do. But the risk is to look at optimizations specific to the niche any given chip is designed to serve and pretend that those optimizations are generally applicable.
In this particular situation: there is no "fast RAM" / "slow RAM" in competing designs; there is only "system RAM" (low end) and "system RAM + graphics RAM" (high end). If you have a single pool, all that really means is balancing contention between the CPU and GPU vs putting the entire GPU subsystem on the other end of a PCIe interface and avoiding contention...
Incidentally, the DDR4 interface (used in the M1) is only marginally faster than a PCIe4 x16...
Good point: a good example is that the Kindle or Audible apps can use content that's bought through the web site(s), but the apps can't direct customers to the appropriate Amazon web page to buy, only the in-app Apple store mechanism.
As to Google, I'm holding no candle for them; if Apple does wrong and Google does the same thing, that doesn't make the Apple action right!
For the Google case, though, I suspect a lawyer could advance a case that while they _require_ Play, there is no explicit prohibition against other methods _as well_. The issue is that the language is fairly non-specific with things like "must use" which is different from "may not use anything except"! In support of this, Google prohibits the use of Play for certain types of transactions, so one could argue that other methods are explicitly contemplated. (Not a lawyer, myself, but I do believe that if an app offered the choice of Play or web based payments, it would be compliant with the text as written!)
The core problem is the duo of high fees and prohibited alternatives, so they literally prevent competition. If the WWDC and all that jazz is _worth_ paying 30% for, then let their developers decide and vote with their feet.
The more I think about the "money laundering" point, the more it seems probable that this is a pure hypothetical designed to justify a particular rate: the notion is that launderers will tolerate 10-15% (or whatever) loss in order to get clean money... so villain A buys 1,000 in-app "things" for $10 each, and villain B collects $10K less the fees, so $8K-$9K. But we're being asked to believe that _all_ villains won't do this if the fee is 30%, which seems a stretch.
Turning around, if the villains won't accept giving up a 30%, why should the non-villain developers...?
Let us be generous and assume that this is true (although one would expect subsequent testimony from the people who allegedly declared that 30% is the necessary level)...
... none of this justifies the prohibition against using other purchasing channels.
(The rest of it is self-serving bullshit -- "we host a very expensive party, so you have to pay for it", "we're building nice digs that some of you will be able to use, so you all have to pay for it", etc -- which might play to a jury, but there is no jury here...)
Just a clarification: a _former_ ATF agent who is willing to testify as an expert witness; this is not the same as the ATF's expert witness!
I'd also note that the issue with the AR's is a _legal_ one, and applying purely _technical_ standards is sometimes problematic (cf tomatoes are vegetables legally, fruits botanically).
In general, courts (especially appellate ones) tend to look at the intent of the law, and it seems entirely possible that, even if the part doesn't perfectly meet the verbiage, it is the best fit, and the bill's introduction allows for some latitude even if the specific language appear not to.
Yes they are, and they are amongst the safest vehicles out there, mostly because of their low center of gravity, but also because of the assistive devices that are standard (and not really any different than those on, say, a high-end BMW, but better than that of a mid-range Ford which leads to the shocker than expensive cars have better safety systems!!).
One of the peculiarities of Tesla is that by choosing to call their assistive systems "autopilot", a bunch of idiots who don't know what an autopilot is think it's an autonomous driving system... which it isn't, any more than the autopilot on a commercial airliner is autonomous. "Autopilots" basically keep a aircraft heading in the right direction at the right altitude (sound familiar?), although of course there are those that can follow a more complex route ("Turn left at Greenland", for example) and, yes, there are other autonomous systems (like Autoland) that can do more.
But exactly like Tesla's autopilot, only a fool would trust an autopilot/autoland system not to put you on a runway just as another aircraft decided to cross (because the ATC voice traffic isn't available to it, it won't "hear" a go-around command).
Unfortunately, there are a lot of fools around, and as there are more cars than airplanes, there are more fools on the roads than in the skies.
Autopilot/cruise control/lane following/predictive braking are all great tools. But the whole point of the report is that it's a very, very long way from that ("Level 2", apparently) to "Level 5".
Now, if Tesla _did_ have something with higher capabilities (say, "Level 3"), why would they put it in a passenger car? What's the value proposition there? Yes, it's nice, but...
On the other hand, putting it in a _semi truck_ makes a lot of sense. Consider a convoy of tractor-trailers following each other nose-to-tail, with the driver in the lead vehicle driving the whole convoy, completely networked amongst each other. Yeah, there are lots of details to consider, but on the long stretches between towns, this seems like a reasonable (if not perfect) idea. Kinda like a train.,..
(It's also worth noting that Tesla has true self-driving cars in the Boring Tunnel system. Of course, this is a controlled environment with only other Teslas to worry about, and they can all be networked, but it's possible that Musk's comment had more to do with this -- public transportation using autonomous vehicles in a closed system -- than anything else).
The idea that multiple vendors would be better value (as is claimed by the AWS/Oracle teams) is, frankly, suspect (as the gov submission notes). Whether or not the admin costs of having more than one supplier are outweighed by the savings on each line item seems less important than the costs of managing twice as many security solutions (and potentially cross-domain solutions between them).
Yes, it's obviously true that no vendor is going to be immune from vulnerabilities, but it's also true that a connection between (say) Azure and AWS will provide twice as big an attack surface.
The only place where the outsourcing _might_ make sense is where the number of staff needed is small; for example, if the premises can justify only (say) 4 hours of cleaning per workday. If you can find and depend on a single part-timer, that's great, but usually it really is more cost-effective to subcontract to someone with multiple people, so they can take holidays or be otherwise unavailable, plus you can get "surge" people for the monthly re-polishing of the anti-static floor tiles or whatever.
(No disagreement with the OP for the large operation, but _sometimes_ the numbers can favor outsourcing).
Have you seen a Soyuz?
A huge difference between (say) Apollo -- where everything had to work right just once -- and Falcon is the proof of the design in the repeatability... It's a good design and flying the thing repeatedly proves it.
(Yes, we lost two Shuttles... but both of those losses came about from failures in the non-reusable pieces: Challenger because of the O-rings in the SRB, and Columbia because of the insulation on the External Tank.)
@Ian Johnston.... that isn't really accurate.
Yes, the USG bought the early Falcon 1 _launches_, but those were the "per copy" costs, not the development bucks. Likewise, NASA paid to have a usable Dragon capsule in operation, but the total amount was $396.... which sounds a lot, but the price of a Falcon 9 launch is $62M, and this was while NASA was paying $56M per seat on Soyuz, so basically NASA paid SpaceX the price of 7 Soyuz seats to get the (uncrewed) Dragon. "Huge funding" isn't really accurate.
Yes, Crewed Dragon was a really significant investment by NASA ($2.6B), but again, compare it against (a) what it would have cost NASA to develop the same capability (see SLS!!!), and (b) what the actual unit costs work out at (because those numbers include a number of launches).
SpaceX also pockets large wodges of cash from the US military, but as suppliers, not development contractors. When they launch a GPS satellite for 70% of the price of the incumbent (ULA) -- which they do -- it's clear this is not the typical government-funded operation.
But here's where it gets interesting: Falcon Heavy was privately funded (hence the payload: Musk's roadster), Starlink is privately funded (1,378 satellites launched and operational to date), and Starship is privately funded.
Sure, they couldn't have got this far without the government's help, but that's true of a lot of businesses (telecomms, IT, aviation, transportation); separating pure private development from development for particular customers is hard and pointless!
Your feeling isn't accurate. What actually happens (and it does in those far-off pre-Musk days) is that groups of people (often called a "village") get a single installation and then distribute the connectivity to a school, community center, etc. And the fee for that single link is either subsidized or paid for by government or NGOs, so the actual out-of-pocket cost to a tribesman in, say, the SaHell is zero but the benefit is achieved.
Back in the "first world", my dad, who lives in the boondocks (almost half an hour from Oxford, so you know it's rural) only got actual broadband a couple of years ago; before that he had ISDN until BT cancelled that and then GEO satellite for a few years, which worked well but had (obviously) a bit of a latency problem. Large parts of Scotland, Wales and Northern/Western England have the same issue: there aren't enough people to justify BT OpenHand or the mobile phone operators building out infrastructure.
I found this snippet from the gummint's announcement interesting:
Their available speeds will rocket to more than 1,000 megabits or one gigabit per second.
So what do we think the plan is? Obviously, we'll be getting more than gigabit, so does that mean they'll be rolling out NBase-T on-premises equipment? Or does each fiber terminate in a box with 2 or more bonded 1000-BaseT interfaces? And as the wording is explicit ("available speeds"), that suggests the more-than-gigabit is net of any additional protocol, voice signals, etc.
Exciting times! Is this the newly non-European UK pulling ahead and quietly promising 2.5G, 5G or even 10G to rural premises?
The alternative, that the DCMS and the not-always-Right Honourable Oliver Dowden may actually be overdosing on their hyperbole and, when they say "more than" they actually mean "almost", is just not credible!!
This is just foolish nonsense. What it actually "just proves" is nothing.
1. Why shouldn't a business be partisan? Explain your answer with particular reference to Fox News!
2. If 81 million people voted for X, and 74 million voted for Y, and you want to reach the greatest number of customers, which group would you prefer?
3. If you are concerned about truth and accuracy, would you try to limit comments from people who maliciously spreads lies and deceit?
4. Your ignorance of the legal system is stunning. As part of this suit, would the defendant be able to conduct discovery on the plaintiff? (hint: three letters, rhymes with Tess)
5. If you're facing an "investigation" by a piece of sh*t politician who has recently sued to prevent another state conducting its own elections in accordance with rulings from its own courts, and who is under indictment for securities fraud and accused by his own office of bribery and abuse of office, do you prefer the idea of a lawsuit managed under the auspices of a federal court or a partisan hatchet job conducted by the aforementioned indicted political piece of sh*t?
"Obviously Twitter is trying to reserve the right to ban people "just because" and "special circumstances" and the AG is calling them on it."
Yup.
Where exactly are you extracting some right to use someone else's property against their wishes?
How much do these people pay Twitter for use of the system? Or Apple, or Google, come to that? So there's no expectation of service, no warranties of availability.
The _only_ reason that Twitter publishes their policies is because they want users, and they calculate that having a fairly light touch in terms of rules.
If this entitled position was even vaguely sane, people could demand air time on Fox News (or OANN or NewsMax or...)
Bizarrely enough, "conservatives" like to pretend they're the party of small government, but they'll happily take from successful businesses on some specious argument that, apparently, someone owes them a website!
It's a statement of opinion based on disclosed facts using an obviously subjective rhetorical label ("far right"), and there's no evidence at all that the clearly non-defamatory statement would or could cause the business any damage.
US courts would dismiss at a motion for summary judgment phase (or earlier, depending on where any putative suit was filed), and while UK courts might actually hear a libel suit (if anyone bothered to file one), there's little chance it would survive any attempt to show damage.
The "disclosed facts" include, conveniently, Amazon's filings in Parler's attempt to sue them: numerous examples of material that contravened Amazon's TOS and which, remarkably, happened to be slanted to the right wing...
Well, yes... but how long does it take to (a) get a hearing and (b) prove to the judge's satisfaction that the discount to the sentence is earned? You _know_ the Arizona Dept of Corrections will produce evidence that their "computer says no", so now the convict needs to prove that the DoC is wrong...
This is a term of art: an uncontained engine failure means that parts of the engine exited the housing. By contrast, a clogged fuel line would result in a contained engine failure.
When UA328 came to a full-stop, parts of the engine were some 16 miles away.
This was an uncontained engine failure!
This is slightly simplistic: engine (and indeed aircraft) maintenance is entirely predictable, and all airlines very carefully schedule their usage to have the aircraft/engine at the right place at the right time, and the "right place" may be a remote station (to you) rather than a home base. American Airlines, for example, have a lot of maintenance done in Tulsa, OK, even though Tulsa isn't what anyone would call a "base"; what they tend to do is fly into Dallas, and then ferry empty to Tulsa; when the overhaul is complete, guess where the check-flight ends up...?
Of course, if you're a small airline, then you care much more (than the big guys) about emergency repairs (if you have two aircraft, and one is broken, you've lost half your fleet/income. If you have 600 aircraft and one is broken, you've got a few hundred cranky customers).
This is mostly false: Texas has two low-capacity interconnects to the Eastern Interconnect (i.e. the grid for the eastern states of the US), and two low-capacity interconnects to Mexico.
Since someone will quibble about the words "low-capacity":
Texas power generation capacity: ~34,000MW
Texas North DC Tie (to Eastern Grid): ~220MW
Texas East DC Tie (to Eastern Grid): ~600MW
Texas Railroad DC Tie (to Mexico): ~300MW
Texas Laredo DC Tie (to Mexico): ~100MW.
Total tie capacity (either importing or exporting): 1,220MW aka 3.6% of the state's generating capacity.
30% of the capacity was offline because of the wrong type of snow...
This is one of the "A" model B777-200s, with a (relatively) short range (basically, good for transatlantic routes, but not transpacific ones). Ranges is a little over 5,000nm, so DEN-HNL at about 3,000nm is no problem.
This particular aircraft is configured for 364 passengers, and there were 231 passengers and 10 crew on board UA328, so not exactly "lightly" loaded, but not operating at max weight by any means!
The B777-200 has a difference of 100,000lbs between max take-off and max landing weight.
UA328 did not dump fuel (total flight time was less than 5 minutes, and they basically just took off, flew in a circle, and landed).
UA328 landed some 30,000lbs overweight on a 12,000 runway (runway 26, as it happens).
This is the second time a United B777-200 has dramatically lost an engine in flight; the other time was 3 years ago, also on a flight to HNL (but from SFO, not DEN). Just like this time, that flight landed safely without further incident. The NTSB finding was that P&W needed to develop a better fan blade inspection. Maybe they need a better better inspection procedure?
Fun fact: UA382 was being operated by line number 5 (the 5th aircraft produced). The replacement that evening that took the passengers to Hawaii was line number 4, which was part of the flight test program (used for things like fuel consumption measurements), so LN 4 was actually delivered after LN 5.
And yeah, LN4 (N773UA) was the aircraft that lost an engine en route to Honolulu back in 2018...
This, this, a thousand times this!
Competent security professionals don't rely on "oh, well, we bought it from someone we like", they rely on monitoring and architecture. Don't trust anyone... its not exceptionally hard to do!
(The hard part is usually figuring out who to trust -- "trust everyone " and "trust no-one" is fairly easy.)
Yep... And remember the last go around had the servers in an Apple-owned data center, and the comment from the Apple IT people about how they'd have noticed traffic -- the alleged traffic -- because they don't put the IPMI/management NICs on the same network as the application NICs, and they monitor their management network.
Which sounds, you know, professional.
True story: I was overseas when I got an urgent call from our cybersecurity consultants... they'd detected traffic from our internal network heading to an Indian IP address. The FBI were alerted. Tension was rising.... until I pointed out that the IP address in question was part of a block that we'd sold a couple of years previously, and in fact this was only an ancient desktop PC trying to reach a long-decommissioned print server whose address was now allocated to an Indian telco!
(And this, people, is why NAT is your friend: using a public IP address range for an internal network is a Bad Idea. See also IPv6...)
Not at all. It's up to them to show that they had reason to believe you really ate babies. Subtle distinction, but they don't need evidence of the eating, just evidence supporting the reporting of the eating ("Reliable source Fred -- who's a spook, so we can't divulge his name -- says you ate babies").
But even worse, this isn't a case of someone accusing you of eating babies, so much as someone saying your babies are being eaten by The Bad Guys. Supermicro isn't being accused of malfeasance, China is (and Supermicro is the victim).
No. It's very hard to prove the "negative" that there _weren't_ spy chips added. And all Bloomberg has to do is say they had reason to believe that there reporting was true, and they're then off the hook even if it was false! Remember, the gist of the story is that The Bad Guys (aka "China") maliciously added the chips without Supermicro's knowledge, so where's Supermicro's beef?
There's a huge difference between a disclosure pursuant to a court order and handing stuff over just because someone is asking for it.
Meanwhile, the judge's comments about how this is an attempt to obtain through the English courts what the Canadian courts won't deliver leaves going... "Hmmm"!
Also... even if that particular PowerPoint can be interpreted in support the idea that Huawei didn't mislead HSBC, it's absolutely certain that there were other documents involved in the relationship, and it's entirely possible that the PowerPoint in isolation was supportive, but the rest of the documents may have been more damning.
Remember, the US _has_ that document, and so if the extradition is successful, it will be provided to the defense at trial _in the USA_. So this is an effort to block extradition, not an effort to prove anything about the underlying charges.
It's also... _odd_ that she / Huawei can't produce their own copy of the blasted thing. Remember, the suggestion is that this PPT shows Huawei didn't mislead HSBC... so why do they need something that HSBC has?
But actually they didn't "cave in France". They created a new product that engages the French publishers. The publishers see a (potential) improved market/channel, and Google retains it's "traffic steering" position.
The problem with the Australian approach is that it's absolute: if Google puts a publisher's article in a search result, they'd have to pay. If someone links an article to their Facebook page, Facebook would have to pay (because FB puts a "preview" of that link in-line).
Leaving aside the specific merits of Google/Facebook's past behavior, I have concerns about this sort of behavior by publishers; where does one draw the line about indexing/cross-referencing.
For example, a citation to a news article in an academic paper is functionally not unlike a search-result cross-reference, and I don't think any reasonable person would think charging for that would be good public policy... but I do know that some publishers would like to have veto power over such citations (e.g. a paper discussing News Corp's egregious behavior, with cites, would likely be unpopular with the Murdochs).
So let's look at a few possible (not necessarily probable) situations: if a publisher gets to charge negotiated fees, how long before negotiations become content-sensitive? Articles critical of a certain political group might be more expensive than those complimentary of them. A publisher in a competitive market could offer free of cheap access to their articles in order to marginalize their competition (and revoke that access once the competitor is suitably diminished). Is this functionally any different from pay-for-ranking schemes, only inverted? Who qualifies as a publisher: if Bruce Dinkum writes a definitive article on how to grill shrimp, does he get to go to Google and demand fees and/or delisting? How about if Bruce instead writes a piece expressing his view that all the brown people should go back to where they came from (unless they came from somewhere he wants to live, in which case they should go back to where they were forcibly removed to)?
As we've seen with the MPAA and RIAA, publishers are hugely prone to try to get legislation that not only protects their walled gardens, but also legalizes guard towers, razor wire and searchlights on the walls. So on balance, even with all their imperfections, I tend to lean towards the Googles/Facebooks/etc when pitted against publishers, not because I think the former are "right", but because I think the public interest is rarely served when the latter control access.
No, Turing's pardon did not stand for anyone else. That's the point.
The 2017 pardon, on the other hand, explicitly covers those who were similarly convicted.
I think perhaps you've missing the bigger point, too: either pardon everyone, or no-one, and whether or not the subject is living or dead is pretty insignificant to the morality of whether or not the pardon is appropriate. In 2017, they pardoned everyone . In 2013, only Turing.
As I noted, though, I think clemency mechanisms are less problematic: because you're old/infirm/a great contributor to society/the breadwinner, it does society less good to e.g. imprison you for years, even if a less admirable individual would receive that sentence.
When Alan Turing was pardoned in 2013, my dad raised his eyebrows and asked... "Why him?".
And that's a damn good question. Not "did he somehow _deserve_ a pardon", but "why didn't everyone else convicted _also_ deserve one"?
Not that either my dad or I disagree with the notion that the whole premise of Turing's conviction was unjust (it obviously was, and he was perfectly happy with the 2017 pardon to everyone else "similarly situated"), nor do we disagree with the reality that Turing did fabulous work which dramatically altered the course of the war. But other people did fabulous work, too (e.g. Barnes Wallis, Watson-Watt, Rowe, Bohr, etc etc)... do they also get a free pardon? If so, what would it cover?
It seems to me that clemency is a much less problematic system: yes, you committed the crime; yes, you deserve punishment; but for whatever reason, that punishment will be reduced.
Odd that you think USB Type-C is "obsolete".
(And the EU didn't, back in 2009, standardize on the Micro B connector; that was the manufacturers who picked that).
Here's the current thinking from the EU: https://op.europa.eu/en/web/eu-law-and-publications/publication-detail/-/publication/c6fadfea-4641-11ea-b81b-01aa75ed71a1
Of course laptops get damaged. My point is that the intersection of (incidents that might damage laptops) and (incidents where the damage was averted by the magnetic connector) is smaller than is often implied.
Yes, it's not an empty set: Magsafe _has_ saved laptops that would otherwise been damaged.
Think about it, though: if the problem was as endemic as is implied, why don't other vendors come up with alternatives? Why hasn't Apple put it on their phones? (I previously mentioned my LG phone with a magnetic USB adapter). Why don't the business laptop vendors (Dell, Lenovo, HPE, etc) address the issue you describe by making the power socket easily replaceable? (I once dabbled with a design from theatre equipment from the 1970s, where the connectors were installed so even if the whole socket got ripped out, it could be replaced and its mounting was built to be enormously tolerant of misalignment).
Conclusion: while the problem exists, it's not terribly significant in the whole scheme of things, just like the problem of spilling coffee into a laptop exists, but it's not so common that there's a significant market for waterproofing mainstream / mass-market keyboards -- even though there are folks whose machines would have been saved by such a thing!
I'd much rather see the cost & effort go into things like less brittle plastic and more metal, etc.
Had it been additive, I'd have agreed with you. But the implementation was truly lousy: it removed /complicated functionality (see the crack about the escape key and vim, which was a real nuisance).
It's also very dated now. Compare with the Asus "Screenpad Plus": an entire second touch screen that you can interact with while not obscuring your main screen. (Asus managed to retain the escape key, too). I'm not totally sold on the Asus solution (I'm a Wacom user, and like the "input" display to the right of the keyboard instead of above it, but that's a personal preference), but it's an interesting implementation!
The Register - Independent news and views for the tech community. Part of Situation Publishing
Biting the hand that feeds IT © 1998–2021