Jon finds promising signs from the (probable) leak of Windows 10X but has a set of final demands before he’s willing to give Microsoft his full backing.
Jon finds promising signs from the (probable) leak of Windows 10X but has a set of final demands before he’s willing to give Microsoft his full backing
Windows 10X is emerging from the shadows. This new version of Windows has been a long time in the making, and it hasn’t been helped by Microsoft’s twisting and turning about what it was going to be.
Let’s first look at the original design intention. The underlying problem that Microsoft has with Windows 10 is both its greatest strength and its biggest problem: the Win32 history. This allows you to run a mind-boggling collection of Windows apps on a Windows 10 machine. Drivers from devices long gone, apps that you bought 20 years ago. Usually it all works. Part of this is the programming APIs that have been common across the decades, but there is also considerable work under the bonnet in Windows 10 to allow it to bend over backwards to support some of the more esoteric quirks and features of older apps. Just dig into the co mpatibility settings for an app to see this in action.
For many, this is an important and valuable feature – especially for corporate IT departments that have historically liked to think in decadelong lifespans.
But it’s created a terrible legacy. It isn’t hard to see why you would want to keep running that app that you bought back in 2010. You purchased a full licence and it does the tasks that you need, so why change it?
The killer word is “support”. Is it possible to keep providing free support for an app that’s a decade old? There are some apps and developers that do just that, and I’ve long been an admirer of VueScan by Hamrick. It’s the benchmark software for driving scanners and, to be blunt, nothing comes close. You can buy a standard licence for £30, which gives you a year of upgrades, or a pro licence for £70 that not only brings much more capabilities but also has lifetime updates and upgrades. I bought my licence nearly 20 years ago and I’m still entitled to free downloads of the latest version. Viewed from that perspective, the initial cost has been lost in the mists of time and is frankly an irrelevance now.
That’s fine when you’re dealing with a company that has a solid future and the determination to provide support, but sometimes you need to keep an app running without the assistance of its creators. Maybe it was written in-house and the original developer has gone. Or it was an outsourced development or purchase and the vendor has vanished. And maybe you are sufficiently locked into the workflow that making changes is a pain point that you just don’t want to confront. And why would you when Microsoft does such a sterling effort at keeping skipware code running for you?
Now turn it on its head. Let’s take the view that apps need updating, that you cannot take a multi-decade view and that everyone, whether they’re a large corporation, an SMB or a home user, has a need to keep code up to date. If you take that view, the necessity to run old apps and drivers washes away. This is fine if you’re dealing with a niche such as scanners, which has a company of the quality of Hamrick to support you – it even provides support for a ton of hardware where the original vendor has walked from the software and driver support. It’s harder in the broader context and only getting more difficult over time.
Then consider the real risks of running old code. If it isn’t patched, there will be a raft of security issues that haven’t been appropriately handled. A lot of malware is Win32 code that works partly because the “let’s support and run everything” position of Windows 10 is also its very own Wild West.
In 2021, there’s much to be said for crying “enough!”, sweeping away the abject mess that is Windows 10 and its Win32 support and demanding something better. Is it any wonder that people naturally gravitate to Chrome OS and iOS/iPadOS?
This is where Windows 10X comes in. Microsoft’s initial conceit was that this would work on multi-screen folding handheld devices, rather like the Surface Duo. However, as we know, this work took much longer than expected, and Microsoft took the decision to embrace Android for Duo, a solution that I still find compelling. Just read Microsoft’s original positioning blog from October 2019, pcpro.link/319blog, which paints a tightly focused solution.
The other problem was the Win32 support in 10X. Despite 10X being, to all intents and purposes, a ground-up rewrite of the entire kernel and stack of Windows, eschewing the historical platform support, there was a strong
“Is it any wonder that people naturally gravitate to Chrome OS and iOS/iPadOS?”
desire to still support Win32 apps. It was decided to provide this through container technology, which allows a Win32 app to run in its own sandbox. This is a bold move, especially when trying to balance the needs of legacy support versus the demand for long battery life on potentially lowpowered devices. This solution appears to have floundered: Microsoft is still saying that it will come, but if it doesn’t arrive in the near future then it’s arguable whether it’s relevant.
Now a build of 10X has leaked and it’s fascinating. Before I go further, I will admit that I haven’t installed it on my own hardware; I don’t let anything from an untrusted source near my hardware and network – even in a virtual machine environment. But let’s just say that a friend of a friend isn’t so paranoid as I am. The design is clean and crisp, and moves all the main icons to the centre of the screen. This is a wise move because the default “shove it to the left” made sense back in 1995 for Windows 95, but is awkward on a handheld touchoriented device.
Apps run full-screen, which makes me so nostalgic for Windows 1.x (the only exception to the full-screen rule is that two apps can run side by side). There’s a big push for integration with OneDrive. Applications are native only from the app store, with no support for legacy Win32 software. Web apps can be wrapped into what looks like a normal app, which is a good move if you take the “fresh start” position seriously. Finally, a Microsoft account will be mandatory: you can’t set this up as a local desktop account.
In essence, this is Chrome OS/ iPadOS from Microsoft. Chrome OS has been doing very well, especially in the educational market, while iPadOS is, of course, a force of nature unto itself.
Windows 10X will only be available preinstalled onto OEM hardware, meaning you can’t install this onto your desktop or laptop computer. Nor do I think this will change in the future: while Microsoft could theoretically offer 10X this as an “upgrade”, this would be fraught with support issues. Would it really want to take on the legacy hardware and driver issues that would appear? No, of course not. It’s entirely right and proper that 10X will be an OEM install only onto new hardware.
The question is whether 10X will be enough. Some have said that
Win32 support, via containers, is a prerequisite for success. After all, if you can’t run your old apps, who will want to use Windows 10X? I’ve dithered back and forth on this issue. I can see the upside to containerised support for Win32 apps but I can also see that it brings huge headaches and overheads. Things only get worse when you have the disconnect of Win32 apps running on containers natively on Intel hardware, but with virtualisation on ARM hardware.
Far better to just say, “this is the new platform, get over it”. Such a robust move requires an equally robust push for developer support, and developers will naturally be reluctant to support yet another platform. They already have Windows 10, Windows 10 on ARM, macOS (Intel and M1), Chrome OS, Android (including Microsoft’s own dual-screen Android platform), iPadOS and iOS on their target list. And making decisions between these isn’t easy.
For me, a key issue will be how strongly Microsoft is seen to be supporting Windows 10X. It can’t end up in the same cul-de-sac that sank Windows RT, along with its billiondollar write-off. I can’t wait to buy a Windows Surface device running 10X that uses ARM in a modern processor. I’m quietly confident that this could be quite special.
Open-source JET
It’s quite remarkable how much code Microsoft is pushing into the opensource public domain. For those of us with long memories that go back to the Steve Ballmer “Redmond Kremlin” era, this still seems rather strange. But it continues to forge ahead. The latest code to land in GitHub is Extensible Storage Engine, otherwise known as JET Blue.
Back in the day, there were two versions of JET, which stood for Joint Engine Technology. JET Red ended up as the storage engine in Microsoft Access and a bunch of technologies that needed that sort of storage. JET Blue was behind Exchange Server,
“For me, a key issue will be how strongly Microsoft is seen to be supporting 10X”
Active Directory and a host of server-side technologies.
I’ll accept that is now probably more of archeological interest to those of us who love to dig into these things. Hopefully it won’t become the IT equivalent of “historical documents” from the wonderful film Galaxy
Quest, and I’m sure there’s much to learn from this if you’re minded to dig through the codebase. Nevertheless, Microsoft’s continuing release of code is to be applauded, whether you’re a disinterested user or a code nerd.
Head to github.com/microsoft/ Extensible-Storage-Engine if you’re curious about JET Blue.
Sad spam
I received an email today. It started like this: “My name is Mr Stephen Mensah, The Branch Manager of a Financial Institution. I got your contact through a reliable source called database through Ghana chamber of commerce.” Of course, I immediately decided that this was genuine and signed up, handing over my bank details to this gentleman.
Okay, I didn’t, but I was surprised by two things. First, that people still think it’s worth sending out this sort of spam. Does it really work? Are people foolish enough to follow up on such obvious nonsense? Second, I’m somewhat surprised it got through my Exchange Server, part of our Office 365 E3 subscriptions.
I must find some time to go through the settings, although I confess this is an experience I rarely want to revisit. Whilst it’s undoubtedly a powerful platform, it’s somewhat user-hostile to the SMB administrator. Maybe I ought to look at regrading my subscriptions to something more suitable to my needs, in the hope that the management experience is improved. If it wasn’t for the bundled Office licences, valid acro ss plat forms and technologies, I would probably have moved elsewhere, but it’s a tough decision to make and pragmatic short-term time pressures always win out over finding the time to make wholesale changes. My solution? Hit the “Mark as spam” button in Outlook and hope that some appropriate magic happens in the background.
M1 drivers
Having talked about the advantages of moving to a new clean platform, I’ve found myself at an impasse. For reasons that are historical, and largely around the need to have lots of high-speed storage, my tape library archiving solution is based around macOS, Thunderbolt, the HP LTO-6 1/8 library and Retrospect. It works well and passes the “can I restore from tape?” test, which obviously is a mandatory requirement. It’s not the only backup and archive solution, of course, but using WORM (write once read many) tapes means I have an off-site copy that can be relied on in court if required.
As part of a reorganisation, I decided to buy a basic-spec M1 Mac mini and a small, cheap LG monitor, just to work as the Retrospect server. At the same time, I decided to upgrade the ATTO Thunderbolt-to-SAS adapter to the latest Thunderbolt 3 version, the SH3128 ( pcpro. link/319atto).
All went smoothly until I tried to download the drivers from ATTO at which point I hit this statement: “ATTO products are not currently supported with the new Apple Silicon M1 MacBook and Mac mini platforms. We are developing a set of universal drivers that will support both Intel and Apple Silicon Macs and hope to release them in the near future. Stay tuned for updates from ATTO on when the new drivers will arrive.”
I reached out to the company but it could give me no timescale for release. I understand the problem ATTO faces: the developer edition Mac mini that Apple made available last summer for ARM development work didn’t have Thunderbolt ports, so it would have been hard for ATTO’s team to do anything meaningful with that. As such, I suspect its work has started with the release of the hardware. In the meantime, however, my shiny new hardware can’t do the job it was bought for.
I have found another vendor
( areca.com.tw) that claims to have M1-supported drivers for its Thunderbolt-to-SAS adapter, but it’s a company I don’t know and have never used. For the moment, I’ll just wait. I’m sure ATTO will get there in the end.
Cloud links from Excel
Here’s an interesting new capability in Excel: you can hook up to live internet-based services (visit pcpro. link/319excel to see an example).
As its proof of concept, Microsoft invites you to “track what you eat with the Food data type”. You put food ingredients into an Excel sheet and then it populates the sheet with information. In this case, pulled from Wolfram Alpha, it can cover calories,
“My solution? Hit the ‘Mark as spam’ button in Outlook and hope that some magic happens”
cholesterol and a huge list of foody items. If you’re an Excel devotee then it’s well worth taking a look at to see what other data types can be pulled in.
Browser benchmarks
There are lies, damn lies and benchmarks. They can be useful, but they’re rarely the whole truth. For instance, I was interested by a recent Twitter thread about Speedometer 2 from browserbench.org specifically comparing the speeds of web-based code on various mobile phones.
Out of curiosity I tried a few devices here. An iPhone 12 Pro gets 183, my daily iPhone 11 Pro Max scored 163. My M1 MacBook Air, meanwhile, topped my personal charts with 212, with my six-month-old Core i9 MacBook Pro with 64GB of RAM sitting at the bottom with 95.
Whilst it would be easy to read too much into this, especially in the relevance of this benchmark to one’s everyday experience of real-world performance, it’s a clear indication that anyone who argues that ARM processors aren’t good enough for real work is espousing a view that’s significantly out of date.
YouTube and Zoom meetings
I’ll make no bones of the fact that I hate videoconferencing over Zoom, Microsoft Teams or group broadcasts by YouTube: you lose all of the nuance that you can have at an in-person, face-to-face meeting.
But this morning, it allowed something that wouldn’t have been possible otherwise, given the ongoing pandemic unpleasantness. A few weeks ago, a friend and senior member of my local village passed away. At the age of 98, Bob was still sharp as a nail, funny beyond belief and wise. I knew him through my local pub, where he was a highly respected elder.
I’ve just watched his funeral service via a YouTube broadcast. This afternoon, there’s a Zoom meeting for a wake. It won’t be anywhere near as good as a meeting of friends inperson, but it’s better than nothing and I’m sure I’ll be able to express my thanks to his son, Rob, when the pubs reopen. Sometimes technology, despite its massive flaws and imperfections, can really help.
time you’ve bought the external box and other mounting gubbins it’s probably not going to work out
that much cheaper, but secondly and perhaps more importantly you’ve instantly thrown away th the warranty on the indoor gateway. gateway
If it goes bang after a couple of months, you’ll be on your own.
Another way people try to sav save money is to buy an internal gatew gateway, but then mount the antenna out outside using a long cable going through a hole in the wall. Again, it’s not to be recommended because LoRa and l long cables don’t really work togethe together. I’ll explain why in a bit.
Once you’ve made a choice of indoor or outdoor gateway, the n next consideration is whether the gat gateway comes with antennas. If not, you’ll need to add your own. These days, most indoor gateways have antennas, but that’s not always the case with external gateways because people like to be able to select their own to maximise the range or else just fit in with the local environment.
You’ll also need to make sure that you buy a gateway that uses the correct frequency band. In the UK, we use the EU standard of 868MHz, and such gateways are often label led EU868. Incidentally, there’s a fairly new 2.4GHz LoRa standard that Semtech (the main chip provider for LoRa) is pushing quite hard right now. It has the advantage of being a single standard for worldwide use. It may well take off and become the main LoRa standard in the future, but at the moment you’ll struggle to find many sensors or other devices available, whereas EU868 kit is relatively plentiful.
The other worry with 2.4GHz LoRa is that it’s fighting for airspace alongside Wi-Fi, Bluetooth and various other devices. That’s not to say there aren’t other users of 868MHz, but they are fewer in number, and tend to be things like weather stations and garage door openers that only transmit occasionally and in short bursts, whereas with 2.4GHz you’ll find lots of continuous use (such as your neighbour streaming Netflix).
For now, I see 2.4GHz LoRa as a novelty, but maybe in a decade’s time it will have become the standard.
The final thing you need to consider when selecting a gateway is the number of channels. The very first devices were single-channel, but they have several disadvantages. The biggest problem is that they, pretty obviously, only listen on a single channel. This means you need to tell your sending nodes to only transmit on that one channel too – and that’s not something that you can adjust on some devices. Unless you have a specific use case, I’d steer clear of single-channel LoRaWAN gateways.
You’ll also see some dual-channel gateways, but these have similar issues and, again, I’d avoid them. In particular, you might come across a cheap two-channel gateway from Heltec called the HT-M00, but this one has an additional problem. If you remember last month, I mentioned that LoRaWAN data packets typically start with a preamble of eight upchirps to allow the receiving end to lock on to the signal. Well, to save costs, the HT-M00 has been engineered such that it needs to see a 16 upchirp preamble. If you use standard LoRaWAN devices with an eight upchirp preamble then the HT-M00 will miss maybe 70 to 80% of the packets. This is mentioned in the Heltec documentation, but in such a way that it’s easy to overlook it.
I argued with Heltec that I don’t think the device should even be called a gateway because there are so many LoRa sensors and nodes out there where it’s simply not possible to change the preamble to be 16 chirps. Heltec told me: “In the
LoRaWAN protocol, users are allowed y the preamble. Those nodes not modify the preamble are blem.” So that told me then! is that with most off-thesors you simply can’t adjust amble length. My advice d be to steer clear of this ice and indeed other single nd dual-channel gateways. What you need to look for s an eight-channel gateway. n fact, strictly speaking, to be RaWAN-compliant a way needs to be an eightel device. Many are available, of the cheapest and easiest to the Things Network Indoor , which is around £75. As you ably guess from the name, door gateway. It’s a simple design, just plugging into a 13A socket (it looks a bit like a Wi-Fi extender), and you can also power it via a USB-C socket, which gives you more flexibility with positioning. It’s designed specifically to work with The Things Network, a worldwide public LoRaWAN network, which I’ll write about next month.
The great outdoors
When it comes to outdoor gateways a good but not stupidly expensive option is the DLOS8 from Dragino.
As I write, it’s £220 at RobotShop ( pcpro.link/319drag), but I can also see it listed for £500-plus from other suppliers. You’ll find this a lot with LoRa and LoRaWAN kit: there can be a massive discrepancy in price between suppliers so it pays to spend five minutes shopping around before you buy anything.
The Dragino DLOS8 is a great introduction, and needless to say it will work indoors too. It has a plastic construction, while most of the competition use metal enclosures, but it seems to be weatherproof. I’ve used one in a position where southwesterly storms hit it straight off the sea, and it has coped just fine with the combination of strong wind, rain and salty water. It comes with a decent antenna too. One useful tip if you plan to set the gateway up and not touch it again for a few years is to seal around where the antenna connects to the gateway either with UV-stable amalgamating tape or, for a slightly less neat finish, a generous squirt of exterior silicone sealant.
These two models are ideal starting points if you want to
“I’ve used a DLOS8 in a position where southwesterly storms hit it off the sea”
experiment with LoRaWAN, but the sky’s the limit if you want more upmarket kit – you’ll find gateways on the market costing many thousands of pounds. However, I’m not convinced that they’re really that much better than these cheaper options.
As we’re operating with extremely low-power radio signals with LoRa, we must be especially careful about things such as antennas and cables.
But, before I get into that, I need to explain a technical term called “link budget”, which is defined as the transmitter power minus the receiver sensitivity. These things are all defined in dBm, which is an absolute unit of power expressed on a logarithmic scale. 0dBm is 1mW (milliwatt), 10dBm is 10mW, but because the scale is logarithmic 20dBm is 100mW, 30dBm is 1,000mW, and so it goes on. Similarly, -10dBm is 0.1mW, -20dBm is 0.01mW. You get the idea.
So if the LoRa transmitter has an output power of 14dBm and the receiver has a sensitivity of -136dBm then the link budget is 14 - (-136), or 150dBm. This means that the total of all of the losses and gains in the signal between the transmitter and the receiver has to be 150dBm or less. In fact, it needs to be a few dBm less because at 150dBm we’d be right on the edge of the receiver sensitivity.
When you consider the path between the receiver and transmitter, there’s a sequence of gains and losses. These are expressed as dB rather than dBm because they don’t relate to an absolute power, or dBi in the case of antenna gain, but thankfully all these different types of dB can be added together. People get quite obsessed about using the correct type of dB, but I find that even such obsessives tend to get it wrong occasionally (please forgive me if you spot one in this column!).
The gains are mainly in the antennas at both the transmitting and receiving end. Those little antennas supplied with the Heltec boards I mentioned earlier will have a gain of around 0dBi, so effectively no gain at all, but a well-designed LoRa antenna can have a gain of between 5 to 8dBi.
Then there are the losses. The biggest loss is what’s known as the “path loss” (or sometimes the “freespace loss”), and that’s simply the attenuation in the air between the two antennas; under ideal conditions, at 868MHz, the path loss is around 91dB per km in air. Of course, as soon as you start to introduce obstacles such as buildings, trees or even weather, the path loss increases.
There are also losses introduced in cables, and sometimes these can be significant. The amount of loss depends on the quality of the cable and the length. So you need to use the lowest loss cable that you can afford, as well as keeping the cable runs as short as possible. Even the connectors will introduce a loss, so don’t plug two 1m cables together for a 2m cable run – use a proper 2m cable instead. And likewise try not to use adapters if the cable and equipment don’t match up properly: make sure that the cable is terminated with a good quality connector that properly fits the equipment.
Let’s work through an example. Assume we have the original 14dBm transmitter and -136dBm receiver, so an overall 150dBm link budget. We’ll work on the basis that the transmitter has a 0dB antenna (no gain), but it’s connected directly to the transmitter so there’s no cable loss. And at the gateway end there’s a 5.5dBi antenna but connected via a cable with 1.5dB loss. Let’s also assume a path loss of 150dB. We end up with:
This is the received signal level, or RSSI. And because it’s slightly higher than the 136dBm sensitivity of the receiver, we’ll be fine. The 4dBm difference between the two is known as the “link margin”.
But what happens if we move that gateway indoors behind a doubleglazed window? That’s probably going to add around two or three more dB of attenuation, at which point the reception becomes very marginal.
It’s the same with cables that you might use to connect the gateway to the antenna. The attenuation varies, but let’s take RG58/U, a typical type of cable used to connect antennas. Its attenuation at 868MHz is around 0.6dB per metre. So, again, you can see that, if you’re working at the extremes of reception, adding a long length of cable to attach an antenna on your roof might work out to be a false economy. It’s far better to put the gateway in your loft and then use a shorter cable.
Well, that’s it for this month, but there will be more LoRa and
LoRaWAN next time.