Posts by rg287

Most people who quit M365 for Google do it out of spite, but there’s no ROI in that

Re: Lesser of the two evils.

It's a tough one. Which do you like least: Microsoft or Google?

Two party system?

I believe I'll vote for a third party candidate!

The analyst said he’s seen some M365 users trying to use the threat of migration as a negotiating tactic when renewing their Microsoft deals. He said that won’t work if you leave it too late, because Microsoft knows it’s not possible to make the move quickly.

The International Criminal Court has entered the chat...

For a given value of quickly of course. But it's amazing what a little motivation can achieve.

Although why people would migrate between Redmond and Google is anyone's guess. Surely openDesk or similar ought to be your destination these days. This may cost more in £ terms than the other two - once you sort hosting and admin staff. Private sector has to make their own risk assessment.

For public sector it should be a no-brainer. Even if the local-hosted solution is more expensive on paper than the MS/Google option, you're going to be funding domestic jobs and skills (training people with actual technical skills, not just as 365 admins) which is good for the UK skill base, and ultimately the private sector and wider economy. Additionally, instead of allowing profits to be slurped off to the US (or their tax haven of choice), you get a portion of your spending back via income and consumption taxes. £120 spent on a local provider instead of £100 with a US cloud vendor is the economically responsible and efficient thing to do - even though Friedman's acolytes will bleat on about being "fiscally irresponsible" because the line item is larger for the UK version.

When the customer also happens to collect tax, it makes sense to prioritise vendors who actually pay tax in your jurisdiction.

_{Oh yeah, and national security and not letting foreign law enforcement read your government's email or something. But that's not important because national security only concerns how many tanks, jets and ships you have. Complaining about having your whole chunks of military logistics infrastructure hosted by foreign cloud providers is nitpicking.}

Switzerland tests a FOSS escape route from Microsoft 365

Re: Remember Munich

It's an interesting case study in how Munich flailed whilst the Gendarme's rollout of Gendbuntu (also a custom distro) went much better - although arguably slower (since they only finished migrating all France's overseas COM-DOMs a couple of years ago).

IIRC they really pushed for evangelists in the user base and prioritised people who were willing to give it a go - you got a new machine and a widescreen monitor (a big deal in 2007!) if you volunteered to move over.

That presumably meant they got early feedback from people who were in a forgiving mood because they had nice kit, before they dropped it on administrators 3 years from retirement who had worked their way over several decades from mechanical typewriters through word processors, into Windows 3.1, Win 95 and were disinclined to learn a new thing just before they left (presumably also meant some of those people could retire out without migrating and you could induct their replacement directly onto Gendbuntu).

Spurs boots VMware, cites 85% licensing saving

Re: 840+ under seat access points

I definitely wouldn't like to be sitting on top of a WiFi access point, if that is what is being described. Far too close to the nether regions. I guess it could act as a seat warmer on a cold winter's day

Low-power non-ionising radiation for 90minutes? You ought to be more concerned about sunscreen and not getting skin cancer.

The problem with wifi in a stadium setting is not so much the bandwidth demand of those users or providing a decent signal but the sheer number of client devices and doing time-sharing over a shared-medium (radio). You could blast out a strong signal from one AP at the end of the stadium to cover everyone, but with xx,000 connected devices, throughput would be pitiful - a bit like standing in central London with full 4G signal but your actual data rate is approaching zero.

Under-seat access points will most likely be dialled right down from their theoretical maximum and run at a lower power than your home router because they're serving a relatively tiny little block of seats. They don't want them to be full-power because they'd overlap and compete with the neighbouring access points (or rather, the APs on the other side of your neighbours - since you'll be alternating channels to reduce interference).

The 5GHz band is extremely useful here since it attenuates more quickly than 2.4GHz did, helping provide extremely localised service in these high-density environments without too much neighbour interference.

Re: Alternatively,

You could always not bother with all that complicated sounding tech and people could maybe do without for two hours and pay attention to the game.

Quite, although to play devil's advocate, aside from internal/stewarding and emergency service provision, a chunk of the complexity and IPTV endpoints are probably for broadcasters and media in any OB galleries, commentary booth(s, the gallery running the big screens, pitch-side networking for camera, monitors for pundits, photographers squirting stills up to Getty/AP in near-real-time, etc, etc.

The wifi install is big, but probably wouldn't be that much smaller if it was only for internal/media/emergency services because you still need the physical coverage in the faraday cage. Serving spectators of course does add some density in the stands (sheer clients per AP) and there would be less pressure to upgrade APs or move to Wifi7 - the bandwidth required for internal services is likely fairly inelastic, so if it works today, it'll work tomorrow. Not the case for spectators trying to pull down 4K streams of other matches!

Apple defies memory shortage with new Mac minis

Re: Yeah I'm baffled....

There is one - and precisely one - properly useful application of Carbon Capture, which is bolting it onto the outflow of high intensity industrial processes that we can't easily do away with like cement or steel plants (SSAB in Sweden has "zero fossil" steel, but I understand it's a fossil-free recycling process, not making virgin steel).

And this is assuming you can't find a direct use for the raw outflow - like British Sugar did when they piped waste CO₂ and heat into the greenhouses at Cornerways Nursery to increase tomato yields.

These chucklefucks spinning carbon capture systems in free air are way down the paradoxical consume-to-save brainrot, particularly since they seem to be building them in places which have abundant renewable energy like Iceland (geothermal) and Kenya (solar), but which don't necessarily have high industrial emissions or especially high ambient CO₂ levels.

The tech is useful for mitigating high point-emissions. Particularly if it can provide a usefully pure supply of food/medical/industrial-grade CO₂, diversifying supply chains (most medical grade CO₂ is a byproduct of ammonia and (bio)ethanol production, which famously had issues in 2018 and 2022. The UK's largest bioethanol plant has also recently shut down, cutting domestic CO₂ supplies by 30%).

Almost nobody pays attention to web standards anymore

Re: Dead Internet Theory

Do you know the name William O’Reilly?

No?

Don who plagiarised his students. And when he was discovered, issued legal threats against the student paper who were calling for a more robust response to his misconduct.

But because his face fits, the media didn’t give a toss.

Arday was most likely a bit of a chancer - though LJM stand by his PhD. But a FoI was made to Cambridge the day he was appointed demanding his credentials. Whilst he had various critics, the most vocal was a chap who had to move back to America after he was (correctly) “cancelled” at Cambridge for saying that if Britain was truly a meritocracy then there would be no black people in public life outside sports and acting. He literally held - in public - that people of colour are the intellectual inferiors of white people. Indefensible.

The Heritage Foundation saw fit to gun for Arday FFS. They have literally no business this side of the Atlantic, never mind commentating on Cambridge professorships.

So maybe Arday was a chancer. He wouldn’t be the first. Questions were raised about the CV of our last Chancellor. Somehow the media circus about a senior member of Government was less frenzied than one about the qualifications of a sociology professor….

Ultimately it was an internal academic issue for Cambridge. The entire media circus is driven by racism and white supremacy. Not always consciously, some of it is just piling on the bandwagon. People deeply underestimate the level to which far-right US-based interests are influencing public discourse through the strategic blowing up of matters like this out of all proportion as wedge issues. It’s a real threat to our democracy and needs to be more carefully scrutinised.

Trump sends the US Navy back to the steam age

Re: "linear induction motor"

Of course, being a metro, it has an average speed of 25-30mph and a theoretical peak of 50mph. And it's a closed metro system with entirely dedicated, grade-separated tracks, so a third rail isn't an issue (no level crossings, and being newish, evacuation walkways were included at build-time. You wouldn't have passengers evacuating along the trackway in an emergency).

If you wanted higher speeds (e.g. mainline rail), you need more power. Getting more Watts to the train means increasing the cross-section of the third rail to handle more Amps, or increasing the voltage. If you increase the voltage then it becomes increasingly difficult to keep it insulated from the ground, which is why OHLE can be energised to 25kV but third rails are <1kV.

SkyTrain is neat, and linear induction can work - and does. But SkyTrain also isn't launching jets off a carrier nor trying to shift mainline Inter-City services at higher speeds. The peak forces and abuse levelled on the system are quite different. It's a good solution for metros. With adequate R&D, it's possible it could work for new-build mainline, but you lose any ability to interoperate with legacy lines that don't have the induction rail and really it's not going to be appreciably better than standard electric traction.

Re: Where's Lewis Page?

Ah well done. You thought I wouldn't look those articles up.

The first one is a lifestyle article about people going back to dating... Lifestyle is always a fluffy section in any paper.

The second one is a round-up of five articles from the week. The full title - which you truncated - is: "Five Great Reads: life under the Taliban, how to fight back against AI, and … aliens"... a round up of 5 articles from the week discussing women in Afghan, the risks around people being emotionally attached to AI, and a bit about the Fermi paradox... Also not an unreasonable fluffy weekend bit.

This is entirely different to sections such as news or consumer affairs - which you would expect to be a bit more rigorous - running ragebait and articles that are technically correct but also horrifically misleading.

Re: Where's Lewis Page?

Guardian has certainly gone downhill since Rusbridger left. Viner is a decent editor, but doesn’t have Rusbridger’s teeth when it comes to the harder hitting investigative work.

Nonetheless, they’re broadly doing sensible journalism - the Graun isn’t running titles like ‘I’ve got more money than I can spend – how can I shelter it from the socialists?’. The Telegraph has gone from “sensible right” to looking like the Daily Mash LARPing as a right wing paper. Some of their material should go back to GOP headquarters and stay there.

Re: Where's Lewis Page?

Lewis writes for the grown-up press these days:

And by “grown-up” we mean “kids grown up, left home and now sliding into dementia”.

I had a look at my gran’s copy of the Telegraph the other day and it seems to have become a parody of a broadsheet. Conspiracy pushers on Facebook are less alarmist than some of the articles. Not Lewis specifically - don’t think I saw anything of his on my skim through, but the publication as a whole… rage bait from cover to cover.

The EM catapults require two crew to operate compared with 12 for steam. They’re objectively better, and you save space in the ship from the steam plant and plumbing to bring steam to the deck.

The US Navy is woefully inefficient - their ships have far more crew per unit displacement than European equivalents because they haven’t had the same recruiting issues and haven’t invested in automation for the surface fleet.

The Ford class carriers are an attempt to modernise operations and cut down on crew. Every crew member you cut also saves a bunk share, mess capacity, medical demand - this compounds when you then get to make your galleys smaller, shrink the cook teams, larders, freezers, heads and showers, etc.

Adding 10crew to a Ford carrier is not consequential, but if the emphasis is to go back towards old-style operation, then the Ford class literally can’t accommodate all the extra crew without stealing space from operational zones for bunk/messing/heads.

Russian missile uses Nvidia AI chip to help target Ukraine

Re: Good news?

It says good (for them) things about their engineering. But indicates that they can’t fab their own (even in small volumes for defence). Which tells you something about the state of Russian tech manufacturing.

Also, the S71 is an air-launched baby cruise missile. It’s cheaper and simpler than conventional cruise missiles. Even though Russia has shoehorned an OTS AI chip in there, it actually indicates that they’re regressing to smaller, lighter, simpler systems - presumably due to pressure on manufacturing the traditional systems.

On the one hand, this is impressive engineering, but it also indicates that they’re regressing in other areas and can’t manufacture certain components and systems, either at scale or even at all.

Elon pledges to give Nvidia a virtual monopoly over the stars

The rockets are really big. A load of stickers would mass quite a lot.

Famously, NASA stopped having the Shuttle's external tank painted white because it added ~300kg. Which is not perhaps that much given that it weighed 726,000kg when loaded, but 300kg on the pad amounts to measurable payload to orbit.

Dev proves LLMs will run on anything – even a $10 microcontroller

GNU Sir pTerry.

Here’s a solution

Yes, I KNOW. But you know what also isn't a chatbot? Semantic search.

The boss is fully in the AI hole.

Demoing Needle made him go "Huh, that's 99% of what we wanted", made him shut the fuck up about spending £££ on servers, vector databases or shoehorning a chatbot into the system. And get this... it's actually useful. It makes the user-experience better! Most search engines have had semantic search for donkey's years, but it was always a taller order for small teams on internal products. Small LMs like Needle or FunctionGemma are easy to train and add a level of user-friendliness to otherwise rigid keyword searches. Peak Google was when they brought in semantic search that would tolerate the odd typo or bit of NPL, but prior to them second-guessing what you wanted. This gives you something akin to that. And you can train it on a laptop and bundle it into an application on quite a modest server.

Even with needle, If people try to chat to our model, it will only fire one of a handful of deterministic tools, or return nothing. It's single-shot, no chat possible. Same as pre-"AI" semantic-search Google.

Whilst some users are trying to treat our search box like GPT, there are also legitimate halfway-house queries that aren't a specific product but mention manufacturer/range or list multiple products. These are not always easy to parse just on keywords. Allowing semantic searches that go "aha, this person has tried to enter three different model names, so we'll kick them to the comparison page and pre-select those models" rather than just shrugging and giving them nothing is neat. Whereas building a vector db and shoehorning a chatbot into it the way some people wanted to... would not have been.

While you may be able to get a small generative AI model running on a microcontroller like an ESP32, you won't get much from the practice beyond dumb simple pride.

If you want a chatbot. But skinnied-down Language Models are also helpful for things like NLP processing in search and doing some very basic decision making/tool-calling. We’ve just been playing with Cactus Needle (26M parameters) to improve an internal search function which would benefit from a couple of smarts over and above the existing keyword search. Particularly because a minority of users are now asking it questions - they seemingly expect any old search box to be capable of hosting a conversation (urgh).

Having a tiny model that is basically just an expensive bit of application code (no GPU, separate LLM server or external service!) which spits out some json that can be passed to deterministic functions can be very helpful. We’re not going to give users a chatbot in this scenario for a variety of reasons, but being able to fire a couple of different handlers for specific types of input and automagically direct them to different places in the platform is neat. Parsing some sort of input to an ESP32 and making some edge decisions on it is more niche, but potentially still useful.

If you really need a chatbot for certain applications, a small tool-calling layer like this can triage input locally and handle simple queries before committing to burning tokens, which is basically the whole schtick for Cactus - “hybrid AI” doing more on-device and only falling back to hosted/cloud services when necessary.

Burnham wants Big Ecommerce to bankroll Britain's pubs

Re: Sometimes.......

0% beer or cider is just piss, not beer/cider.

Weird how it still tastes better than full-strength Carling and Budweiser though...

Re: Pretty much all costs are simply too high in the UK

Yeah, Amazon didn't kill the high street. Out-of-town retail killed the high street. The town where I grew up built a bypass (much-needed) and the new Tesco/B&Q/Argos destroyed the centre of town. There are a hard core of people who say "well what was the council supposed to do? You can't stop progress". Which is a bit strange because they absolutely could have just kept that land zoned for agriculture and not retail. Indeed, not 10 miles away a neighbouring town managed a similar situation much better, keeping the new Sainsbury's inside the bypass with a walkway to the high street. The car park was a municipal pay-and-display, not Sainsbury's-only. They had geography on their side - being in a bit of a valley which naturally constrained sprawl and limited what developers could ask for, but still.

There's this very curious mindset in many places where if a national brand "offers to invest in the town" then you have to roll over and lay out the red carpet for them. Fuck the 25 boring little indy shops and cafes who are going to see their footfall collapse. "But look how many jobs <brand> has created!", Yeah, and how many jobs/businesses did they destroy? The only difference is that now the profits are being sucked back to the mother ship instead of retained locally.

Re: ecommerce providers are taxed to an extent they're on or near parity to retail shops

Expecting the E-tailers to pay an amount proportionally equivalent to the rates etc paid by the brick and mortar retailers based on turnover/revenue is not unprecedented.

Add a multiplier to business rates that is inverse to the employees per sq.m? So office units, hospitality and retail (i.e. "normal working spaces") would be okay. Vast roboticised warehouses would get it in the neck.

Given that the best estimate of Amazon UK's effective tax rate is ~7% (compared with the 19-25% that indy retailers pay if they're silly enough not to funnel profits through a loss-making Luxembourg subsidiary), this seems entirely reasonable to my mind.

Re: 2 separate problems

Drink driving and off-licence laws just aided that decline.

Drink driving wasn't a problem up until the mid-1970s because:

1. Most people didn't own a car

2. Most people's boozer was walking distance (along with work, at least a couple of shops and the bus station). So even if you did, you probably walked to the pub.

Since widespread car ownership led to the construction of soulless edge-of-town cookie-cutter housing estates with no character or amenities, you more or less have to get in the car to find a boozer, which then led to drink driving, which we obviously had to ban.

eBay is just a neutral platform. All the retailers there compete equally (of course you can buy ads, etc).

Sure, Amazon started as a retailer, but they are now a platform and retailer, with the retailer having privileged access to platform data for every other retailer on there.

Also, eBay don't care if you sell product on your own website or on other platforms cheaper than you list it on eBay.

Re: 2 separate problems

They will die because their purpose has been legislated out of existence

Their purpose was twofold - as travellers rests (i.e. Inns) and as a "third place" before you had village halls or community centres.

They are arguably at their most important today as a "Third Place". If you had a few incels going out to a pub and meeting people who aren't 4chan once in a while the world would be a better place.

But of course you need to be able to afford to go - which is hard when there's been a decade-long real terms devaluation of wages, fiscal drag, inflation-busting rises to costs-of-living (namely housing & energy), etc, etc.

Aren't they currently in the process of trying to reduce their clientele count?

This is clearly how the franchise wars start!

Amazon needs more than taxing, it needs breaking up.

Unlike eBay, Amazon is both platform and retailer. The retailer side uses sales data from independent sellers on the platform side to track trending products and work out what their next Amazon Basics line ought to be. The platform will also cheerily promote Basics products next to products from independent sellers, regardless of reviews or sales performance.

NYT 2018, Reuters 2021

In any normal world this would result in indy sellers abandoning Amazon as a sales platform because of this flagrant anti-competitive behaviour.

But because Amazon is so wildly dominant as both a monopolist and monopsonist, you can't afford not to be on there. But once you are there, you're bound into their junk fees and contractually giving them favoured-nation status, meaning you can't sell your products cheaper than you do on Amazon - not even on your own website doing your own fulfilment.

Amazon needs to go at least four ways - Sales Platform, Retailer, Prime Video and AWS. Not just different boards or "internal firewalls". Forced divestiture out of the control of Jeff Bezos for 3/4 of the business units.

If we don't, then we just end up in a corpocracy where Amazon is the only retailer (and all restaurants are Taco Bell!).

Google meets the neighbors and gets both barrels over its new UK datacenter

Re: Complainants backgrounds......

I thought all or nearly all UK railroads were without level crossings,

Lol.

Inter-city mainlines mostly, branch lines (even quite large ones), no.

It is policy to remove them as and when major renewals allow, and they certainly aren't building any new ones, but in some places this is extremely difficult - e.g. where the line runs right through a village and there's no way to grade-separate the railway from roads because of butted up housing. The only options would be a road bridge higher than the adjacent housing, or to close the line for a year and fully cut-and-cover it. There's a section of West Coast Mainline in Staffordshire that has at least 4 (IIRC) level crossings in the space of a couple of miles (running through Stone and Barlaston). Notionally a 110mph line - but when I used to work around there some trains didn't look like they're were making double digits!

The tories were at least consistent in trying to eradicate UK manufacturing.

This one is a significant Tory donor. The exception that proves the rule perhaps.

rather than useful suggestions more trees, turning lights off after a certain time

Eh, they're not even going to do the easy things - even though they could.

A well known UK manufacturing firm built a large factory within eyeline of my parent's house (although they've got a couple of miles separation). The plans showed lots of planting along the embankments that were raised around the perimeter. In theory you shouldn't have been able to see any part of the building once it was mature. But they never did, and the council certainly weren't going to make them. You can't see the factory from the road going past it because they used the foundation spoil to raise earthworks as a screen, but of course you can see it from further back (i.e. most of their neighbours).

There's loads of light pollution because the exterior floodlights are all above the height of the perimeter berm.

All commercial lighting is fundamentally facing the wrong way. You have big floodlights on the building facing out, because it's cheap to install them there (bolted to the wall, you already have power in the building). The reality ought to be that whether you turn the lights off at night shouldn't be an issue because you should mount any required floods at the perimeter facing in. No risk of locals having light beaming through their bedroom windows then.

But they won't because that costs a bit of money - putting in a buried cable round the perimeter, raising lights on poles instead of on the building. And obviously we're all well aware that in a £300m build, fitting lighting in a fashion that doesn't ruin life for your neighbours is the straw that will break the camel's back. Everything must be done in the cheapest way possible so you can book record profits and the chairman can have a new Gulfstream.

UK.gov's tech department gets the chop after two years

Re: DCMS

I can see why you'd stay Anon!

When GDS was launched there was a lot of fawning media - "these wizards who will revolutionise government with slick data integrations and dashboards", etc.

Yes, bollocks.

But GDS have also been setting a lot of sensible pacing and moderation, and had an important role in data.gov.uk, along with standardising on open formats like .ods instead of .xls. Which is a start.

When everyone was mithering about "But we need an app" circa 2012, GDS were moderating with "What does your website look like? If you can't run a mobile-friendly, accessible website then you're not ready to be jumping on the app bandwagon". What do your users actually need? Spend resource on a better site rather than an app which will stop working in 18 months".

GOV.UK is a fucking delight because it is boringly conservative - a predictable and static layout, doesn't change it's JS framework every five minutes. Minimises tech support for elderly parents trying to work out where to tax their car this year because it's the same as it was last year. And the Cabinet Office was the correct place to run it from - the fact that DWP were running DirectGov was always weird given that it led off to lots of other services under DVLA, etc.

As of January 2025 of course, "GDS" became a different beast because the old GDS got rolled in with the Geospatial Commission, AI Incubator and CDDO, all under the GDS banner.

The fact that Government is heavily entrenched in Microsoft is not strictly GDS' fault, given the GDS remit is online services rather than desktop/hardware/productivity, which ultimately rolls up to Crown Commercial Services (who signed the current £9Bn SPA24 arrangement with MS).

Re: ironically they were making a change

in the name of selling a host of proprietary solutions which almost work

Like AI?

Re: DCMS

Except that GDS hasn’t gone back under the Cabinet Office, which is really where it ought to be given its strategic role in setting cross-Government policy. Being lumped into DCMS seems like a sideline.

And of course “Digital” being in with CMS was always a stupid thing given the remit is distinctly more Businessey. It only ended up as DCMS to start with because Matt “I’ve got an app” Hancock lobbied to get it h set his culture brief.

For the most part, just regular reshuffle stuff, but GDS isn’t really in the right place.

The truth nobody wants to admit: Chinese or not, open models are competitive now

Re: Open Weights or Nothing

that a 3tn parameter model is runnable on a "small little network" is pushing the boundaries of that description pretty far

It's not a box in the corner of the office, sure. But the guidelines are ~64 accelerators (B200 or similar) to achieve sensible inferencing performance for production users. You can easily do that in a single rack (most such servers are 2.5-4 accelerators per OU, so 16-24OU), provided you can get 150kW in (and suitable cooling).

This is a big old line item, but also not hyperscale and well within the reach of PLCs/enterprises (who also might just rent time from neoclouds). In the context of the sort of business that can afford such a rack, it's probably a pretty noddy little network compared with their broader server fleet, office networks, production/manufacturing facilities, etc. If you're in the business of running moderate computational clusters for CAD/CFD/Modelling, then this will be a very expensive rack (given the RAM density/cost of the accelerator cards) but not really large or complex.

HS2's latest reset ditches autonomous train tech to get project back on track

Re: feasibility testing

How exactly do they determine that the higher speed is not feasible?

Not through crashing trains. The question of proving the trainsets is more of assessing the sweet spot of speed; comfort and maintenance/wear. It's no different to doing acceptance tests/sea trials of a new warship. You shake it down and check everything is working as intended. The problem we have with HS2 is that there isn't a test track in Europe that can go all the way to the higher speeds - so we would have to trial the rolling stock on HS2, which would then incur delays if problems are found with the rolling stock. Normally you would have already certified the rolling stock by the time the track is ready for opening. This is why they're proposing to certify the stock to lower speeds, open at lower speeds, and they can then run the acceptance trials and tests over the next year or so and increase the speeds.

We know the higher speeds are perfectly possible - the rail speed record is 357mph, set by a French TGV on the (then) new LGV Est in 2007. To achieve this they had to adjust the tension on the overhead wires, raised the voltage (31kV up from 25kV), used a specially prepared train with larger wheelsets and a few other tweaks. You could run passenger services at that speed but it would eat the infrastructure and be immensely maintenance intensive/expensive. So we can do it. We just don't.

As an aside, they achieved a speed of 225mph in 1981 and 320mph in 1990. So when people whinge about "Oh, but 220mph would be the fastest revenue service in the world, it's gold-plated speculative risk", like yes, most lines still run 180-190mph, but people have been studying this for 40 years. They just don't get to build new lines very often.

HS2 is using slab track, which is newer tech (but well proven) rather than the traditional ballast on LGV. Slab gives better dimensional rigidity and lower maintenance costs than ballast, so there's no real question that HS2 could support a train running at 300mph+ provided that the alignment is straight enough (which it is). We just don't want to pay for those tickets or have it constantly closed for maintenance.

Re: Vanity

Labour proposed it, by the time the Tory/Lib Dem coalition came about and green lit the project they didn't have much choice given how much had already been done and spent on it.

What is this drivel.

* The Tories started talking about HSR at their 2008 party conference.

* Labour were a bit cool on it, but in 2009 Brown commissioned a Command Paper which was delivered after the 2010 election to Cameron/Clegg.

* Cameron absolutely could have buried it - it was a pre-viability study. The idea they "didn't have much choice" is an outright lie. Anyway, all the major parties approved of it and committed to moving it forwards.

* In 2019 Johnson commissioned the Oakervee Review, with the hope it would recommend that costs had spiralled and everything was too expensive. Inconveniently, the review concluded that it was absolutely necessary and cancelling HS2 at that late stage would create all sorts of supply-chain issues for suppliers who had already started kitting up with contracts in hand. The idea of "we'll spend it on the existing lines" was identified as nonsense which would take many years and cost more for less.

* In 2021 Johnson cancelled HS2 to Leeds anyway, because he hates devolution and anything north of the M4.

* Sunak later cancelled the Birmingham leg stating that "the facts have changed" without being drawn on which facts. He then announced "Network North" which has been cooked up the night before by his SPADS in their hotel. The DfT knew nothing about it and they basically just listed out a bunch of projects which had already been funded.

Re: Vanity

One of the major ideas behind HS2 was actually to turn Birmingham into a dormitory city for people working in London because house prices in London have become so expensive.

A major idea that didn't make it into the original paper High Speed Rail (commissioned by Brown, delivered to Cameron), nor was recognised in the Oakervee Review.

I mean, you might be right... but it is a conspiracy theory because there's no evidence of it outside Stop HS2's plaintive cries of "But nobody cares about getting to London 20 minutes quicker".

And even if some nefarious brandy-swilling illuminati types do hold that view, the requirement to unblock existing lines for local and regional services is undeniable and true. Any argument against HS2 is also an argument against having built the motorways in the 1970s.

There have already been a large number of developments of rentable apartments within close proximity to Curzon Street Station and more are planned.

Yup. This is what happens if you announce new infrastructure - you get inward investment, much as Stratford saw huge private investment when the government built out a load of infrastructure to support the Olympic Park. What's interesting is that in many other countries, the rail operators are encouraged to buy up and develop areas around new stations, capturing some of that increased land value (that is a direct result of taxpayer investment). This is very sensible - if a business is going to locate in an office because of it's proximity to a major station, it is better that their rent goes to the rail operator to subsidise fare and operations (the value of that office is partly because of the money government has spent). But in the UK we let those profits get privatised.

This is one of the points Burnham has been making about building/funding HS2/NPR/other infrastructure - you can partly fund it through land value capture by examining the pre- and post-investment land values and applying a modest tax to the difference.

Re: Japan's Maglev is slightly cheaper!

And it still doesn't work.

China built their Shanghai Maglev to much fanfare but it now operates to 200mph to save on maintenance costs because at 300mph it was eating itself alive.

And Maglev can't switch onto the existing railway to continue north (or integrate to HS1 in future).

Maglev is a truly point-to-point system which can't play nice as part of a larger system. If we had a saturated High Speed line (as Japan does from Tokyo to Osaka) then maglev might be a natural progression, but you need your conventional network for local/regional travel, high speed for IC travel and then maglev is only worth layering across the top when the HSR network is straining.

Re: Vanity

The current journey time from Warrington or Manchester to London is 2.05 hours.

The current journey from Manchester to London blows through little mothballed stations like Wedgwood and Barlaston without stopping. Because of the demand for inter-city services, local services have been cancelled in their favour, despite demand and a thousand new houses being built next to Wedgwood.

The entire point of HS2 is to remove those non-stop services that blight towns and villages to no local benefit off onto their own line (the equivalent of a motorway) so that locals can have local services back. The current situation is like someone driving from Manchester to Birmingham nad going through the centre of Wilmslow, Congleton, Stone, Cannock, etc instead of using the M6.

But you're happy with your 2hr trip to London, so fuck the rest of us plebs.

Building a railway for 225mph is a lot more difficult than for 125mph as you have to go more of less in straight lines or at least very large radius curves.

So... if we eventually hook it up to HS1 (which will happen), you're happy for Manchester-Amsterdam services to crawl along at 125mph before getting to London and putting the hammer down and doing the 180mph that HS1 is capable of. On what planet do fifth columnists like yourself live?

Worse yet, you want to build a wiggly line which locks it to 125mph forever - even if you operated at 125mph on opening day, obviously you build it to a minimum of 190mph - the thing the French have been doing since the 1970s (LGV Sud-Est opened 1981 doing 190mph), so that you have the scope to upgrade in future.

Another issue is that the HS2 trains won't fit in a standard platform, the coaches are too long for standard curved platform and the wrong height.

Get a clue. They do. From day one, the trains will pull off at Lichfield and stop at places like Stafford and Stoke on their way to Manchester. There was initially an idea of a mixed fleet of HS2-only and "Classic Compatible" trainsets but they've now standardised on a single class.

Re: HS 2 late

They could have delivered much of the passenger capacity increase by sorting out WCML congestion into Euston

Nope. Get bent. Downvoted. You do not make statements like that without offering a sensible proposal for how. Otherwise it's just FUD and "Oh no, but what about this other hyperloop thing we could do.".

Network Rail are not stupid. They make their money from track access charges.

If they could improve capacity to Euston they would - but it's rammed. Platform capacity is at a premium.

And you're more broadly wrong as well. Sorting out WCML congestion into Euston does not fix platform constraints at Birmingham New Street or Manchester Picadilly. Curzon Street will let them literally triple the number of local cross-city services, which Brum desperately needs (currently there are long distance trains parked on through-platforms at BNS for 30mins doing crew changes, which block the possibility for local/regional through-services. BNS was never supposed to be a terminus for long-distance services).

That being said post Covid, with WFH … has traffic even returned to normal baseline levels yet ??

Yes. The mainline traffic bounced back by the end of 2022, albeit with long-distance leisure making up for a lot of commuting.

Nationally, overall ridership was sat at 98% of pre-COVID for a long time, but this was because London commuting has remained down and did not reflect wider trends. A bit like saying average (mean) wealth has gone up but really it's just a couple of billionaires getting richer and the median is static. It's a function of bad stats.

Many areas have seen ridership bounce to 140% of pre-COVID levels. We need express trains off the legacy network to make space for local and regional services.

Re: HS 2 late

Even worse, the speed promise is what has made teh project so damnably expensive.

At those speeds they need it to run in a straight line as much as possible, as I understand it.

This is basically irrelevant. Running it anywhere in the Home Counties is going to be expensive. A straight line means you use less land than a (longer) wiggly route, but does also mean you sometimes clip an expensive bit of property. These two things somewhat balance out. Also, you'd have to be an absolute de Pfeffel Johnson to build a brand new Inter-City railway in the 21st century and not build as straight as possible. Conventional HS stock will do 180-200 mph, and that's what the track will be laid for, but they've very sensibly bought and civil-engineered the alignment to be good for 250+mph so that in 50 years as material science moves along, you can upgrade with new overhead lines and rail tech and run at high speeds - without having to buy bits of land to straighten out the bends that would be left if you half-arsed it in the first place. Passive provisioning is a good thing. We know from the West Coast Route Modernisation that torturing higher speeds out of old track is far more expensive, disruptive than new build, and basically not worth doing. Electrifying and fixing knackered bridges and stuff, sure. But trying to get a 80mph line to 120mph to avoid building a new Inter-City Route is a way to madness. Anyone who disagrees with this presumably issues developer laptops with 4-8GB of RAM and instructs users to come back and ask for 16GB when they really need it. It's 2026, not 1826.

The major cost centres of HS2 are:

* The stations. In France, stations are built by the cities. A new LGV project would actually be scoped from the equivalent of the M25 to Birmingham International. The new stations and approaches would be scoped separately. So when people take Project Cost/Route KM and come up with a much larger number than the French equivalent, it's because it's an apples-to-steak comparison.

* Constant spec changes - started building and then politicians got their crayons out. de Pfeffel is personally responsible for £1.5Bn of waste when he tore up the Euston Station drawings and decided he knew better than the rail engineers.

* HS2 Ltd. Why the fuck they set up a special-purpose Ltd company instead of just having Network Rail be the commissioning body is beyond me.

* Treasury Intransigence - They have this weird desire to smooth spending as if they don't have a central bank at their beck and call. You want to build infrastructure as quick as possible in order to start realising the economic gains of that outlay as quickly as possible. Slowing down and "smoothing" spending costs more overall and delays the economic gains.

* NIMBYs - The Acts of Parliament authorising the route didn't authorise enabling works. So HS2 Ltd have been stuck in a position where they have an Act of Parliament saying "You can build a railway through this bit of land", but they've then had to apply for planning to build an access road or groundwater monitoring station - which local councils have refused. This - perversely - lengthens the disruption and uncertainty for local residents because you've got half a railway built through, but certain works are artificially lengthened and a subset of councillors feel smug about their petty "act of resistance" that - overall - changes nothing.

* Skills shortage. Because we do start-stop development and then the workers disperse to other industries, it's required a huge training and apprentice programmes. Countries that do rolling programmes don't have this issue because they just keep rolling hard-won skills and experience forward instead of having to reinvent the wheel for every new project.

Notion kills its Gmail client after AI agents keep humans from troubling inbox

Re: Notions Use

I can imagine (trying to?) use Notions' AI as an email re-distributor.

More likely connecting it to JIRA/whatever and using it to parse an email, open a ticket and add relevant tags or assign suggested support bods depending on the contents. It'd probably work passably in a lot of cases, and tickets can be reassigned if it gets it wrong.

Email -> Ticket is not new, but required someone to triage the new tickets.

How the AI bubble could pop and take down the global economy, according to the BIS

Re: To be...

While there are some very important applications that AI is being successfully used for it is being grossly over hyped in many others.

Machine and Statistical Learning are being successfully used for some very important applications.

The stochastic parrots are not. Don't need them, don't want them and the frontier researchers like AMI have already abandoned them as a dead end in favour of World Models.

I stopped believing in article’s from the BBC at the turn of the millennium.

Articles from the BBC definitely exist.

End of era as the BBC switches off Radio 4 Long Wave service

Re: Irreplacable...

I am not sure what the ongoing cost is, but it does feel like we have lost something reassuring and important

And it seems to be indicative of the direction of travel.

There is currently muttering in the air about when FreeView and FreeSat will be decommissioned in favour of Freely (Freeview but streamed, which means they can start carrying 4K content, or allowing Freeview SD channels to output an HD stream even if they can't get spectrum for an HD broadcast). The speculation is mid-2030s.

This has three problems:

1. Freely is crap. We updated our (decade old) TV just before Christmas and after tuning the new unit in, it then asked if we wanted Freely and we naively said yes. The guide takes forever to load, channels can take a while to start streaming. It lies that it has no internet access when it's wired to the router and the YouTube app works. We did a hard reset and tuned it back in, declining Freely. I suppose it will come. But the viewing experience just keeps declining. Analogue channel switching was instant, digital introduced a slight lag and IP is just slow.

2. Presumably it only lasts as long as the Freely software gets updated. If the backend servers migrate to different codecs or APIs in 8 years time, the TV stops working. You can fall back to using the tuner for broadcast TV, if there are still broadcasts to tune into.

3. You require an internet connection. Which is a thing that costs money. Now yes, you're supposed to have a TV licence to watch Free-to-Air telly, but as it stands, anyone with receiving kit can tune into the six o'clock news. If you need a network connection with a minimum bandwidth as a prerequisite for your receiving equipment to work, then this is a non-trivial additional burden, particularly in adverse weather conditions or something where local infrastructure may have broken down. There have always been remote areas and valleys that couldn't get terrestrial TV, and Freesat sorted out that group. If you go to IPTV, then they may be stuck again.

On that third point, you also screw a new and different set of people. My parents are in the West Midlands. Well populated area with decent telecoms and infrastructure coverage. But their phone line comes from the opposite village to the one they are nearest. In 2026 they still struggle to get 8Mb. They can get FreeView, but Freely? Nah. Streaming is not something they are able to do.

The other - 4th - point, since Digital Sovereignty is popular, is that Freely requires not only an internet connection but DNS. DNS is controlled by ICANN. If the US ahem "leadership" went mental and did something nutty with DNS overnight, then you lose your national "broadcast" infrastructure and also your ability to update end units (e.g. with hard-coded IPs, etc). If the internet disappears, the UK retains a sovereign broadcast capability. But if all the public-information/broadcast comms are routed over the public internet... you're stuffed. Yes, that's very extreme, but we're allowed to hypothesise extreme scenarios when we talk about "How does the Prime Minister address the nation in a national emergency?". Not over the internet is the answer.

In a more apocalyptic scenario, then with broadcast, anyone with a receiving set and power can tune in. But streamed, you require your hosting, IXPs, ISPs and all the other interim service providers and kit to be powered and up.

Notwithstanding the specific challenges of replacing the Long-Wave hardware, we really ought to be keeping ourselves some sort of radio-wave broadcast system for radio and TV. If the worst happens, anyone with a solar panel and battery and at least receive, even if London and Manchester have been nuked, and all our IP capacity obliterated with them. In the scheme of government spending, it's far less than the cost of running and crewing one destroyer a year, and arguably as important to our national defence and resilience. National security is not just about how many tanks, guns and planes we have. Most nations are 3-6 square meals away from anarchy. Food security, energy security, robust telecoms... all apparently "nice to haves".

GOV.UK goes Dutch on payments as it dumps Stripe

Guess the government wanted a change.

Whilst UK Gov will presumably have the volume to be on undisclosed custom pricing, Stripe's "Standard" card rate is far more expensive than Adyen's.

Adyen have a fixed price structure for stuff like Direct Debit and Open payments, which GOV.UK Pay may have calculated is better for their specific transaction profile/amounts than Stripe's %-based pricing.

Standard card transactions:

Stripe: 20p + 1.5%

Adyen: 11p + 0.6% + Interchange (~0.3-0.4%), so typically 11p + 1%

Winner: Adyen

-------------------------------

BACS Direct Debit:

Stripe: 1% (min 20p, capped at £4)

Adyen: 66p

Winner: Adyen for transactions >£66

-------------------------------

PaybyBank:

Stripe: 20p + 0.5%

Adyen: 11p + €0.50

Winner: Depends on transaction value profile.

So based on standard pricing, local authorities using GOV.UK Pay/Adyen will basically halve their card payment costs. They will win if they're taking a lot of Direct Debits over £66 (which council tax will usually be). PaybyBank is a bit more variable - Adyen becomes cheaper on payments over £80.

They're also adding a bit of resistance to US sanctions and intransigence. For instance, Karim Khan will be able to pay their council tax. Sanctions blocked his use of Visa/MC (we need a UK card payment network). But to compound it, Stripe LLC will also enforce those sanctions, which then blocks PaybyBank and DD, even though neither method in itself has a US dependency. Going Dutch is still not optimal, but short of nationalising WorldPay (London-based, US-owned), it'd be a bit of work to set up our own or convince banks to do a sovereign solution.

2,000 retired Google Pixel phones get a second life as a private cloud

You think making use of stuff that would otherwise end up in landfill is 'greenwashing'?

Not OP, but kinda yeah. To do this you have to strip out the main board and acquire a custom-made power and Ethernet adaptor (which presumably then needs some fiddly soldering and tinkering - how many mobile boards come with Ethernet breakout pins?! I suppose it’s being attached via the USB controller?).

So the reuse requires some virgin manufacturing. And then we’re sending the screen, battery, etc to recycling.

As others have said, better that Google supports phones for longer, or provides an unlocked boot loader for the last update so the phones themselves can continue to be used.

This is not so much reuse as recycling-with-one-component-being-diverted-for-reuse.

They do mention the main board makes up a disproportionate amount of the embodied carbon, so that’s something. I question what the actual Watts/Operation comes out at. We all know that old-generation hardware is less energy efficient, which is tolerable for periodic/client usage but not as good for always-on applications like servers. It’s why old supercomputers get retired rather than donated or handed down to lesser institutions - the cost of power (and the room to put them in) is too high. By the time they’re being retired you can often get the same performance from half as many racks or less.

On a small scale this doesn’t matter since the whole cluster can probably be powered from a couple of solar panels and turned down at night.

The point they make is valid - there’s a tonne of low-computer functionality needed and a tonne of mobile devices with decent compute getting churned out of use.

But the emphasis ought to be on life extension for handsets, which is partially limited by battery replaceability but also by software updates and locked bootloaders - which is 100% in the control of Google.

Signal says UK plan to scan devices for nude images 'endangers us all'

Re: Zombie Starmer's Victorian internet: An Internet without Things.

I've run across smut from Victorian times.

The Victorians were a horny bunch in their own way. Far from the usual prudish, repressed stereotype. But made up for it by overcompensating in public and being a bit obsessed with policing outsiders and sexual rebels.

A bit like how the reddest, most homophobic US states end up searching for the most gay porn when Pornhub release their years in review data.

Palantir wins £9M contract to run UK firearms licensing: CIA-backed biz to hold gun, bomb, and poison records

Re: Hypocrisy on Steriods

Woooosh.

The point being made up there is that the required software is not particularly complex. It’s a basic CRUD database to track the progression of applications/renewals, the assets held by each certificate holder and a couple of integrations/flags for things like medical records. It’s sensitive data, but not a difficult application to build.

And in answer to your rhetoric, bobody ever died because of the quality of FLMS.

In terms of mass shootings, we’ve fortunately only had 4 in the UK (not statistically significant), of which three were Police human incompetence/error. Plymouth was down to the licensing office being “a dangerous shambles” (in the coroners’s words), Dunblane similarly saw CSP eviscerated in Lord Cullen’s inquiry, with multiple attempts by officers to revoke the perpetrator’s license overruled by management. Similar concerns were raised ahead of Hungerford and not acted on.

Weird how they don’t have these issues in Europe, despite much higher rates of gun ownership <mutters something about post-Thatcher policing and public services>

The only one that the Police can’t be blamed for is Whitehaven, there a certificate holder ran into depression and went postal. It’s unlikely there’s much way to foresee that (even if you further tightened regs from where they are now, the people that need firearms vocationally - farmers - are also the ones most likely to have a quiet breakdown in the privacy of their farm).

Interestingly, Lord Cullen’s inquiry into Dunblane didn’t recommend banning pistols. Said there would be no point and at most you could ban individual ownership but people should still be able to go to clubs and shoot. Blair banned the lot anyway. What Cullen DID recommend was training for Firearm Licensing Officers. Blair DIDN’T do that because it’s boring and doesn’t get headlines like bans do. Thirty years later, in the wake of Plymouth, there is a prototype training course being rolled out.

Don’t ever believe that the UK firearms licensing system is robust or a “gold standard”. It looks good on the surface in the same way Reform promise the world. But it’s cobbled together from knee jerk reactions and feels. There are gaping holes and years of labour lost in make-work that has no public safety benefit but sounded good to someone in Whitehall during a crisis. Palantir can’t fix that - the workflow is baked into the statutory basis of licensing. The way “slots” and one-for-one variations work, the distinctions between Section 1 and 2 Certificates. It’s not been meaningfully updated since 1968.

Europe built sovereign clouds to escape US control. Then forgot about the processors

The problem with that is in a data Center, you may need something like ME to manage a machine even if it is running. Bear in mind a data Center may have hundreds if not thousand of servers, so it might not be viable to have a human visit every single machine.

In a data centre, that's the point of out-of-band management like iDRAC and iLO, which can perform reboots, provide remote console and poke around the hardware including power control - but has it's own NIC and sits on an out-of-band network which (in a DC) should probably be air-gapped from the internet.

IME/PSP is of more use in client devices where it can integrate with vPro Fleet Services and MS InTune.

ME/PSP can provide IP KVM and other niceties, but in a data centre, you arguably don't need them (except that these days they do more than just management, and are integrated with pre-boot integrity/TPM type roles, so Intel/AMD would have to rearchitect elements of their chips to do away with them entirely). It would make the sovereignty issue cleaner if Xeon/Epyc chips just came without, with management done via a discrete OOB board on it's own network.

Tony Blair and the Reform party both want AI in government, though they can't agree on what for

Re: He sees good use for AI...

He's just a salesman for Ellison:

And his son. As was pointed out by Kevin Beaumont yesterday, Euan Blair runs an edtech company called Multiverse that now has GenAI ambitions. They acquired an AI startup called Searchlight in 2024.

On the first reading of Blair's essay I got strong impressions of "Davos groupie helps billionaires feel better about themselves".

Then I saw Kevin's post on Mastodon and realised Blair is also just shilling his son's grift. His efforts to shoehorn AI into government and campaign for lax AI regulation all makes a great deal of sense in that context.

The basic thesis of his essay was correct - Labour have no "project" or strategic ambition. They have policies but no coherent plan. You can see this in transport where they've announced things like Northern Powerhouse Rail, but not as part of a wider high speed project. Whereas the original NPR used elements of HS2 and was integrated.

This is fucking stupid. Even if you can't build everything all at once, you want to have a big masterplan that shows how the elements of your system fit together. When we built the motorways, the government didn't just say "right, there's some money for councils to build strategic roads. Crack on". The MoT drew up a masterplan of what a UK strategic road network looked like, and County Councils then delivered roads in their region based on that plan. The US Interstate Highway network was bult along broadly similar principles - states building roads to a federal set of plans so that their new infrastructure all joined up at the borders!

Unfortunately having stated the obvious (Starmer has policies, but no broader vision or ambition/plan), Blair's prescriptions are standard Davos apologism - cut taxes for the wealthy, cut planning regs, drill-baby-drill in the North Sea to get cheap energy for AI datacentres (he hasn't worked out that the reason we haven't exploited those North Sea gas reserves is that whilst they're quite sizeable, they're also hard to get at, and only financially viable if gas prices remain high. So the idea of prioritising cheap energy over clean energy by increasing north sea production is oxymoronic. North Sea gas is increasingly expensive. And gas prices are now likely to drop since the UAE left OPEC and wants to produce more. Cheap energy means Arab energy - or renewables.

He even gets shirty in his essay that "people complaining 40 years of neoliberalism have achieved nothing presumably includes the last Labour government".

Errr, yes it does! Stoke schools left to crumble by PFI contractor, IPPR estimates that NHS will pay £80bn for £13Bn in private investment (a standard mortgage or business loan would have been cheaper!).

Blair's failure to have National Grid (plc, bizarrely) build out enough capacity and interconnects has got us where we are, spending money on turning down highland wind power because the grid is saturated north of Edinburgh.

UK begins antitrust inquiry into Microsoft's business software ecosystem

Re: Charity

As a small charity we would not get volunteers to use the online tools unless they were "Microsoft that we learnt at school."

Have an OnlyOffice instance ready to go. Next time Microsoft change M365 (again), switch over and say "Oh yeah, this is what it looks like now. Blame MS. I know, it's so annoying how they just keep changing stuff".

Only half joking.