Posts by Charlie Clark 14449 publicly visible posts • joined 16 Apr 2007
Loads of the other apps have traffic data: Here We Go does, as does TomTom. If you use any of these services, you have to remember that you're also providing telemetry data for them.
Where I think Google may excel is the use of different routing options when the standard topological ones don't work because it logs so many journeys rather than just where the vehicles are at any one moment (or where they may be in 30 minutes).
On my phone I use "Here We Go" for the car and OSMAnd for walking and cycling. Both have offline maps with frequent updates. Here We Go benefits from traffic information from all the car manufacturers that use it. Most people I know seem to use Google Maps but I can't get on with it.
The traffic here is often so bad that it makes no point in starting down the tunnel of alternatives which can quickly lead you angry, stressed and disoriented. But local knowledge helps most: driving to France last year and, even on a Sunday, I knew it would be best to avoid Paris but I didn't insist upon it. Didn't pay attention quickly enough to the first warning about traffic on the Périphique and ended up driving, well SWMBO was doing the driving, through Saint Denis and other parts. It did save us a few minutes but it was also quite hair-raising! Needless to say, the return journey was via another route!
It depends very much what you want to do and the less frequently you do something, the less important it is what the running costs are. I'm more than happy with an inkjet for the 4 or 5 pages of colour I do a year but duplex is great as is the ability to print on A3.
China is already leading the way on LTE development but has also understood how to play the interoperability game with the rest of the world better than America ever has. Yes, with a large enough domestic market you can launch your own technologies on the world but adoption is unlikely. The problems with TV signals (NTSC, PAL, SECAM) led Europeans to understanding that only common standards would have a chance in the future and this was the reason for GSM, designed to get companies to work together. While various competing standards may initially have had advantages, over time it was GSM specifications that got the most traction with Japan, Korea and even America adopting them over time.
China understands committees and the needs for standards and has been a player from the start. It could choose to introduce something like CCCP (Chinese Communist Party Communications) and might be able to sell this to poorer countries with little international travel. But it's simply easier to bring specifications to the committee and win the race to implementation.
America's history in international standards business has, however, on the whole been miserable, with various companies championing various competing standards that they hope to enforce via patents on an unsuspecting world.
Used correctly, machine learning can be an excellent tool in medicine and is already is use for processing many routine scans, where it is better and faster than humans, but they have the last say. But it's not possible to create one without significant access to reliable data and domain expertise, which you're not going to get from somebody's phone or strava account.
I think the battle about delivery confirmation was lost decades ago and greylisting has long been an accepted practice.
As for charging to send, it's clear who this would really benefit, but it would be almost be impossible to implement worldwide. DKIM and DMARC are great but what's the point when spam filters don't prioritise them?
Not quite true: Microsoft recently tightened its spam filters, catching a lot of legitimate stuff that previous went through. For example, e-mails to me from the BSI (German Office of IT Security) are being held silently in quarantine, despite having correct DKIM headers (SPF was always a joke) from well-established servers. Mind you, Mailbox has decided that all Apple Developer e-mails are spam and my ticket about this is in its third month.
To make things worse, the rules are stricter for new accounts and must be overruled per account. GMail is similar as my initial e-mails to be people tend to land in their spam. :-/
I don't see this as a race to the bottom at all. The Chinese models are bringing much needed competition to the market. Furthermore, they also have other paying customers in vertical industries such as autonomous vehicle services which they're already exporting around the world. And, as with many services, it's going to be industrial customers and not the "prosumers" that the media are currently fixated about that will drive the market by only paying what they think services are worth.
"It's intended to have lithography mask production, and then logic, memory and packaging, and chip testing all under one roof," he said. "So you can have a very fast iterative cycle." Musk acknowledged it was a "high-risk, high-payoff bet on AI chips."
Hm, I wonder why no-one else thought of that? Probably because experience shows that it's not a good idea. Not least for maintenance purposes. Oh, and where is all the water going to come from?
I don't thinki it's a coincidence that CPU and memory manufacturing diverged years ago. Some people, who know more about the business than I do, say that much of the cycle isn't profitable which is why TSMC and others (well, their governments) are effectively subsidising cheaper chips for us all.
Pretty sure these have been bought in from some Chinese manufacturer. The costs of setting up a production line for what is going to be a couple of thousand units at mosts would never make sense in the US, or pretty much anywhere else for that matter.
So, in the accounting I bet this gets booked under "promotional materials and activities".
Part of the "Donroe doctrine", apparently:
Scott Bessent, the treasury secretary, recently described as “economic statecraft in the 21st century”, in which global access to the dollar and the American economy is “no longer unconditional”
From The Economist last week: the rules-based order is over, you do what you're told or we'll cut you out.
Personally, I'd call this economic and diplomatic lunacy by treating friends and allies as potential adversaries…
I think it's worth adding that the competition within any particular industry is fiercest within China itself. Yes, subsidies do play a part in gaining market share for exports, but they don't explain the incredible pace of development in fields such as telecommunications,s batteries, electric vehicles, solar cells and more recently semiconductor manufacturing and, of course, LLMs. The CCP has so far tried in vain to intervene, because the lack of profitability does carry risks, though the same could be said of the US.
And trying to enforce some kind of ban open source is going to be about as successful as Canute's advisers suggesting he could command the tide. Though, I'm sure this is something that would appeal to Trumpty Dumpty…
In the real world the price comparison may be somewhat less impressive and speed for "real" tasks tends to be the determining factor for many. However, this will still "good enough" for many to want to pay either on their own hardware or somwhere else to run it. And this is despite all the handicaps that the Chinese developers are working against: limited hardware options and active restrictions in some cases. However, it could be that, as in evolution, it's precisely these restrictions that will make them outcompete. We're now starting to see the first systems that can use hardware optimisations on Huawei silicon. Again, China is generations behind both in software developmen and fab process, but it is iterating faster.
It's a mistake to see recent increases in market share by Chinese companies driven solely by subsidies and lower costs. If we don't admit that they have outcompeted us in many areas, we will never catch up.
The marginal price for inference is pretty much the price of electricity: China has been investing more in its grid over the last 20 years than America. It's not there yet, but it now does have some huge (even bigger than anything in Texas) wind and solar farms out west that could soon be plugged in with close to zero marginal cost. Providing the models as open weights provides added incentives to "try before you buy" – China doesn't really care because it knows the next generation of models are already in development.
I think that's reasonable to argue that Chinese subsidies are competing with those offered by the US finance industry. For example, the rules for including the various AI companies in stockmarket indices have been relaxed. This virtually guarantees that tracking funds will be forced to buy the stock, effectively subvertint the market and providing a strong incentive for VC firms to keep giving the companies money, knowing they've got a guaranteed repayment when the IPO happens. This kind of funding has led to some of the circular investments, again signs of market disfunction, and preferential deals with utilities: you pay more for electricity because the data centres down the road got good deals.
But Chinese governments are no longer subsidising these companies as much anymore because they don't need to. Competition in China is so strong that they coined a term for it involution. This is the environment in which many Chinese sectors are operating and only the most effective companies will survive. Silicon Valley doesn't build companies for this kind of environment, instead it provides funding for companies in the hope that winner takes all and that, where it's not possible to beat the competition, you can just buy it. China is aware of this risk and has already vetoed the sale of Moonshot to Meta.
I'm not sure, because if that was the idea, it failed because it was a Chinese model that fixed the problems: 10 points to team Xi. Not for the first time does this make me think that the Chinese understand the technology, the opportunities and the risks better than the money men of Silicon Valley and Wall St.
No, I don't. When I started talking to them, there were no other "outsiders". Committee members said that there had initially been some engagement but this had fallen off. I assume the relevent people no longer had the time or had moved on to other things. My own experience is that work on specifications is exhausting and one of the reasons why a couple of the issues I submitted are still open. It doesn't help, of course, that reports are managed as MS Word files… and response from MS can take a while. So, even when there were monthly meetings of the Working Group, many issues would still take over a year to resolve. In addition, there is a mismatch between working as a developer and seeking clarity about the specification, which is what most of my engagement was about, and the work of a standards body. There are also no test suites for any of the specifications, especially things like the EBNF used in things like number formatting and headers and footers. Still feedback from MS was generally usable, I was able to get a change in Excel over a bug related to sheetname.
And, frankly, some of the bugs in the ouput of LibreOffice were simply embarassing. I think much of the code was written when reverse engineering the old BIFF formats, of which OOXML is often just an XML implementation, and entirely avoidable when working from the specification, for which Microsoft even provided tools.
More recently, the lack of engagement from other developers is effectively allowing Microsoft to extend the specification with blobs – newer charts in OOXML are often just a couple of lines with all details being kept within the app.
I've worked with the specification in detail and, apart from the transitional/strict divide, which use completely different namespaces, there are lots of areas in the specification where things are not defined. These include, for example, how formulae work in data filters, especially whether they apply to text or numbers.
I've worked with Microsoft to clarify some areas and, to be fair, they do have people you can talk to, but the answer is often "that is up to the application". Nevertheless, the only way to improve things is to keep talking to the ISO Working Group, something that the Document Foundation stopped doing years ago.
I think national ID cards – common in most countries – are where to start. Cheaper and easier to explain the benefits for the holders, such as, for example, making it valid for travel in and out of the Schengen area. This whole digital ID stuff fell at the first hurdle because it was all about the benefits to others. Optional digital functionality can be added later.
Anyway, quick win for Burnham, so why not take it.
Just as Dutch and English are partly based on Germanic dialects but they very much are not German.
This is factually incorrect, but a common mistake when looking at evolutions. While it's arguable that Dutch is close to some German dialects, it is distinct enough to be considered a separate language and cannot be readily understood by Germans.
But Anglo-Saxon diverged from continental forms 1500 years ago and was subsequently much more heavily influenced by Latin, old Danish, Norman French, etc.
But, back to the subject: you get a real feeling for Darwin if you boot into single user mode. As for "modern", much of that is a matter of taste or fashions of the time (plists), but the MACH kernel and Objective C were definitely good ideas.
I'm glad to see the flight was aborted before going boom. Something SpaceX has been a bit cavalier about in the past. I'd like to say that, as a listed company, it's likely to be subject to better governance, but Musk still holds most of the shares and nearly all of the voting rights. You might wonder why your pension fund handed over money you entrusted to it with no element of control, but nice animations of happy astronauts playing on the Moon and Mars should keep you happy.
The deadline was set by Trumpty Dumpty so that the mission would be while he's still in office. No-one involved in the programme thinks the deadline is remotely possible and he doesn't really care either. He likes to make announcements about all kinds of things and pretend that the announcements themselves are proof of the fact.
Yes, I've lived with a few apps requiring new API levels but the previous versions tend to run fine.
My 2020 MBP will be 8 years old in 2028. Normally this is when things like the battery start to fail – I've had to replace a few Mac batteries in my time – and some of the daughterboards are a bit iffy. But I'll still have things for it to do.
As for Apple's security updates – the company still has a very poor record in this area. It relies a little too heavily on the good design of the OS, obfuscation and a low attack surface: why go after a few million Millon when there are hundreds of millions of PCs and billions of phones out there.
This statement literally is tech sovereignty is concentrated in the hands of a few global leaders. If power is concentrated in a few companies, then there is effectively no sovereignty.
Europe certainly has the skills to migrate away from US-cloud systems, and the courts may soon start requiring this since the US Supreme Court declared supposedly independent agencies subject to government control, meaning all safe harbour agreements are null and void.
Start with what will cause the least disruption: mail servers and data storage.
The point about the Chromium bug reports is how they're treated and how quickly patches are available. Quite a few of the bugs in tracker are from their own tests and the fuzzer.
And I think it's a step from this to saying the programmers are shit; the same goes in fact for Microsoft's programmers. Lots of code that was state of the art at the time, turned out to be susceptible to attacks that weren't possible at the time. So, again, it's not about how many bugs there are, but how you deal with them.
Apple still has a dismal record when it comes to bug reports, including 0-day. It deliberately decided to eschew the BSD / unix approach of providing patches as soon as they became available. Sequoia hasn't received and security updates this month, though it's difficult to imagine there aren't exploits out there.
Integration within the OS would have been easy enough and most people would rarely notice because for many things, you're just stopping and restarting services, so system restarts could remain fairly infrequent. Credit has to go to Steve Jobs and the team at Next for doing some excellent work on the kernel and BSD userland to make this all happen, but the I-Phone and the desire to turn all computers into consumer electronic devices went to his head.
Some of the code will be very old and from the time before such practices were really practicale let alone established: back in the day it either worked or it didn't.
However, things really weren't helped when Microsoft sacked its entire testing team. It didn't integrate into the relevant development teams or change the process, it just sacked them because it was the easiest way to lower the headcount and juice the numbers at the time. It hasn't really recovered since then.
Google rightly takes a lot of flak for some of its decisions – though withdrawing free services doesn't really count – but it has for a long time had clear (if a bit weird) coding practices, including testing and dedicated security team that actually looks for bugs. That's why we do read about the occasional SNAFU in deployment, but also see Chromium bugs fixed very quickly. I'd argue that much of this stems from the fact the Google uses a lot of open source software and even employs developers to maintain open source projects, where testing has long been standard as has gratefully accepting good bug reports. Yes, Chrome does do a lot of tracking, but the underlying Chromium is there for those that care.
Fundamentally, Microsoft hasn't changed in how it sees customers and, therefore, how it develops software: embrace, extend, extinguish still seems to be the strategy, with extinguish also applying to any kind of quality assurance. And, as long as people are happy to use Edge as a browser, who can really blame them?
In theory, the Windows index should be incremental… but it's always been a resource pig with poor control.
I think MacOS does a pretty good job, though I have had it fail a couple of times recently, and it should by default not index external drives instead of this requiring some arcane command line magic.
Of course, there are still plenty of occasions when I still want to use find.
Surely, the IPO was all about having enough money to own the infrastructure to keep everyone's data safe?
You mean it wasn't?
Mind you, given what he done over the last few years to ruin any trust people had in his companies, it's a wonder anyone is using the AI services. Anything that runs Grok is going to make the simpering robots of the Sirius Cybernetics Corporation want to run and hide! Now more doors wishing you would have a nice day but barking at you like Gauleiters!
I think the CAPEX point isn't made often enough and it gets worse if you start to add charging infrastructue. We've got used to, possibly artificially cheap, Chinese panels and the vestiges of feed-in tariffs, that we think this is how things are always going to be. But if you don't have a use for the power you're producing and can't sell it, then the economics can start to look very different. Of course, car parks themselves are an example of how extremely wasteful cars are…
Yes, it does. Panels are around 25% efficient with optimal placement both of the panels and the sun, you've then got the transmission losses associated with low voltage DC lines – how often do you switch up to high voltage AC? The economics for deploying degrade fairly quickly when you don't have this, especially when you're to sell the power. While the Chinese have driven down the unit cost of panels, this is partly due subsidies and the real costs of making the silicon cells along with the silver and copper you need to connect them will not go away and you'll need lots of optimisers to handle the problems of variable sunlight.
This is why, for photovoltaic production, you really can't beat south-facing sites in spaces with low economic value for predictable yields. I'm sure some wonk could come up with the equivalent area for a convenient mountainside to really illustrate the difference over time.
We're also increasingly hitting problems of excess production at midday in the summer with no long-term storage and no cheap short-term storage.
I think you miss the point – you only need to build the clone once.
The main economic for open source has always been that proprietary software overcharges for the value added. This is why most companies that take open source software private usually struggle to make money with it.
With Windows, it's long been debated as to whether Microsoft actually makes money from the licences – when you think about the amount of work they have to do to keep it patched – but, of course, it's really gateway to selling much more lucrative Office and Exchange services, which are also much harder to swap out.
