Posts by Decade

Re: What is systemd

For the most positively biased introduction to systemd, I suppose there's Lennart Poettering's seminal blog entry, Rethinking PID 1. His entire blog is like a stream of consciousness of the current state and intended future of systemd.

For a whole lot of tedious technical details, there's the freedesktop.org site, with links to download the source and the documentation. Yes, they do have an entire section about the canonical way to write the name of systemd. A lot of the documentation is just linking back to Lennart's blog, though.

If you want to hack it, you need to keep in mind the principles of the systemd cabal: POSIX is obsolete and cross-platform is irrelevant, so don't bother with it. Backwards-compatibility is harmful, so don't worry about the past. Use and depend on the features of the newest Linux kernel.

Re: How Widespread?

MacBook Pro - i5, OSX Yosemite, plenty of RAM.

I've taken to launching Chrome only rarely as needed. I find that Chrome makes the system less stable; for example, randomly, control panels that require password input would not unlock until I quit Chrome. Why does Chrome interfere with System Preferences? I don't know. It's frustrating. And playing video in Chrome runs both cores at 100%, draining the battery very quickly and making the UI very unresponsive.

On the other hand, right now, Chrome in fullscreen on a rotated screen on Mac does correct subpixel antialiasing. No other program does this. Chrome in desktop on the rotated screen also doesn't do correct subpixel antialiasing. WTH, Apple?

I've had bad experiences with Chrome on Windows, too. My favorite period was back in May, when Google Drive crashed Google Chrome. I had to use Chrome for Flash and Firefox for Google Drive. (I also no longer install Adobe's Flash Player separately. Flash sucks.)

Did you read the same article I did?

I read the paper in IEEE Spectrum, and I think it says something very different than what Lewis Page seems to think.

The PhD engineers didn't say that renewable energy was insufficient for producing electricity. They were saying that even if we could wave a magic wand and turn all electricity production into renewables, we've already exceeded the “safety threshold” of CO₂ concentrations. We exceeded that limit about 25 years ago. And besides electricity, there are plenty of uses of energy that are not anywhere close to ready to be replaced by renewable energy, so at best RE electricity will just slow down the increase of CO₂.

Their point is that renewable energy is necessary, but not sufficient, to halt the climate change. We need to focus on the bigger picture, and fund the research and development to fix it.

Firefox is deliberately slow

I don't use Safari on my Mac because I don't know about other browsers. I use Safari because nothing else works as well.

Firefox's problem is that it's not native to anything. The XUL toolkit was supposed to make cross-platform development easier, but it's slow and not native anywhere. These days, computers are fast enough to run XUL decently, but back when it was introduced with Netscape Communicator 6... It was part of why Netscape lost all its market share.

And then development has issues. I use Firefox on my Android because I loathe Google Chrome, but I put up with a keyboard glitch for almost a year, because it was doing something nonstandard with input and that's how long the fix took from being identified to being in the stable release.

Now I've stopped using Firefox on the desktop, because it can't handle multiple monitors correctly on MacOS X 10.9 Mavericks, and now I've installed MacOS X 10.10 Yosemite and multiple monitor support is unusably worse. Considering how long it took for the keyboard glitch to be removed from the Android version, I'm not holding my breath waiting for the desktop version to be fixed.

Stupid PDFs

These days, I mostly use the PDF readers built into Firefox, Chrome, and Safari. Sometimes I use (Apple) Preview. This is on a Mac, obviously.

What I'd like to do is banish them to an untrusted AppVM, as in Qubes OS, but I'm rather addicted to my computer having performance. Maybe next time I build a computer.

Re: WD or Seagate? Hmmmm....

As useful as Backblaze's data are, they're using the drives in a completely different way than I am. They're buying the drives in bulk, and they're using proprietary replication software to stay ahead of hardware failures. I can't afford to buy drives in bulk, so I need each drive to be as reliable as possible.

In particular, I'm noticing that they do not have Toshiba drives. I don't know how reliable Toshiba drives are. HGST drives are more reliable, but their per-TB costs are much higher than Seagate and Western Digital. Digital storage technology sucks.

Seagate is big in arrays... Because you need RAID for reliability

3 out of the last 4 hard drive failures that I've experienced were Seagate drives, and the drives aren't that old, but I don't have a lot of drives and that could be a statistical fluke.

For real data, you need a large population, like what Backblaze has done. For Backblaze's usage, Seagate is still economical because their per-TB costs are low enough to cover the failures, but I don't have that luxury.

Hitachi is now the reliable-drives arm of Western Digital. There are only 3 hard drive manufacturers left on this planet: Seagate, Western Digital, and Toshiba.

I think hard drives suck, but all digital storage technologies suck, so the only reasonable option is to have multiple copies with automatic backup and verification. Like RAID-Z with snapshots and replicas.

Metcalfe's Law will disagree

Bob Metcalfe (inventor of Ethernet) says the value of a network will increase exponentially with the number of nodes.

Walking across the room to switch on the heat is not a terrible burden, and we’ve had thermostats to do that for decades. (It would be nicer if the thermostat had useful interaction with the seasons. 21° is refreshingly cold when it’s 27° outside, but 21° is excessively hot when it’s 16°.) It’s a bigger problem when you’re managing a commercial building and need to walk up multiple flights of stairs to reach all the thermostats. And you pretty much need the Internet if you’re managing an entire campus (Microsoft’s 88 Acres, which is not online anymore).

For me, the matter isn’t whether it’s online (online is better), but whether I have control over where the data go. That sort of security is crucial and currently missing from the Internet of Things.

Bill Gates cleverness

Bill Gates in the 1980's and 1990's absolutely thought of himself as being clever. I shouldn't be spending time on citing the sources, so I won't...

Bill Gates claimed that Microsoft was going to beat IBM because Microsoft was run by smart 20-somethings. He wanted to surround himself with more smart 20-somethings when his mind inevitably declined with age. Of course, when he did reach 40, he changed his mind.

Bill Gates liked people to know about the times he would take a week off to do some intensive reading before announcing major strategies for Microsoft. His most important memo being the Internet memo of 1995.

The culmination was when he wrote The Road Ahead, which Bill Gates used to try to shape the future of technology. It didn't turn out much like the book.

I think the real turning point was during the antitrust case, when Bill Gates decided that he was too special to show up in court, but had to deliver his deposition in a video call where he visibly appeared to be dismissive. Though, I wonder if it were not actually very clever, because he managed to enrage the judge so much that the judge was taken off the case.

Steve Jobs didn't do PowerPC

Steve Jobs did not approve that partnership. That was John Sculley, the guy who fired Steve Jobs.

Steve sure was happy to sell the PowerPC as long as it was viable, but I suspect that he never really committed to it. There were persistent rumors that Apple had a lab running MacOS X on x86, and MacOS X never really ran smoothly on PowerPC. As soon as IBM stumbled with the G5 and Intel recovered from the Pentium 4 stumble with the Core series, Steve switched the whole Mac line to x86.

Re: With you there Tim

The falling costs are a major part of it. My organization doesn't even track assets below a certain size. I haven't asked the beancounters what the exact number was. So, a $10,000 PC in the 1980's would probably find itself on an asset sheet, but a $700 convertible laptop would not.

Noooo.....

I really liked OpenFirmware back in the day. No need for the peripheral controller to have binaries for every architecture, because they were programmed in Forth, compiled to portable FCode.

ACPI is horrible and should be destroyed. On x86, it works because everybody's given up on doing things the right way and just tried to be bug-compatible with Windows. I read a rumor some time back, that the ACPI committee deliberately rejects any good ideas from outside the committee, so ACPI is a complete design disaster in every way, according to no less a luminary than Linus Torvalds.

Mark Shuttleworth also considers ACPI a security risk.

Who can forget Smalltalk? (Everybody. It was a long time ago and nobody used it.)

I just saw an oblique reference on folklore.org, but Smalltalk was a graphical environment where everything on the screen was an object, and you could trigger a menu on any object to display its code. You didn't have to dig through the page's code to find the object of your interest.

I was sad when the object-oriented dreams of the 1980's degenerated to the stream-of-text model of Unix and Windows. And Java is a very poor imitation of an object-oriented language. We're gradually rebuilding, but it's so slow compared to those days, and it's an uncomfortable fit on the Unix base.

High tech does not mean high speed Internet

Here in San Francisco, so far Google, Monkeybrains, and Sonic.net have tried and failed to bring affordable high-speed Internet to the City. The only high-speed option is Comcast, with a normal speed in the 10's of Mbps. In my work place, Comcast was not available, so we switched to Sonic.net's bonded ADSL2+ service, which runs over AT&T's lines.

In this area, AT&T and Comcast are the most common options. Not far away from the urban areas, like Robert X. Cringely's house, the options dwindle to about 0. And in some random parts of the Bay Area, gigabit fiber is available. It's a mess.

Re: Wide open?

The open guest network is not that bad. Presumably the owner would not be using the guest network, because that's tedious. And guests aren't normally an everyday occurrence. You'd pretty much have to lie in wait, biding your time, until a guest comes and enters the password. If it's traffic you want, it's much more efficient to go to the nearest coffee shop.

Much more fun are the crappy Cisco routers where the supposedly "secure" network is vulnerable to the Reaver attack.

Re: Free Software is a requirement

Personally, IoT without Free Software is impossible, because I won't buy it.

Trade secrets? Excluded. Patents, especially on the software? Of dubious benefit to society and should be eliminated. If you want me to buy the thing, then I need to be in control of it.

Because, while they still use the OpenSSL library, they need the early disclosure to prepare packages for their own users.

Also, being developed in the open, LibreSSL is doubtless already being installed in production systems somewhere.

Re: Ride the City (www.ridethecity.com)

I'm planning a route, and I know that a couple miles of it have a moderate incline that will tire my riding companion. Google says the route is "mostly flat" and doesn't give any elevation data. Ride the City says there's an "elevation gain" of 190 ft, but doesn't say where the gain will be.

Re: @ Andrew Orlowski

Sometimes, Andrew, I wonder why you're paid for what you do.

4G, WiMAX and 5G should be frightening the crap out of the incumbents. Why aren't they?

Mostly, it's because the incumbents have most of the power in wireless. AT&T and Verizon have the most spectrum, as sold by Congress and regulated by the FCC. It's difficult for a competitor to arise. Wireless is a shared medium, so you aren't going to stuff many household-Netflix-worths of traffic through it. I recently tried WiMAX, and the latency is through the roof, and there is a lot more packet loss than wired.

The carriers welcome wireless. It's a way for them to claim that there is competitive broadband. But unlike real broadband, where you really have to work to hit the multi-hundred-GB bandwidth cap, wireless has a cap of like 0.5 GB to 2 GB. A 4G connection lets you use an entire month's allotment of data in less than a day. If you use more data, you pay dearly. And if they can get Congress to allow them to let go of wireline, then they will happily stop maintaining the wires and force everybody onto cell phones, like a third world country.

Re: "...grand turnaround plan from Elop that was supposed to save Nokia."

What a revisionist story.

When Elop took control of Nokia, Samsung wasn't even considered a threat. The main reason not to go Android was the very legitimate fear of losing their investment in Navteq. The main reason to go Windows was bizarre (that Nokia would be so big they could influence Microsoft's development), and as it turned out, the naysayers were right. Microsoft's slow development of Windows Phone has hurt Nokia's desirability and their attempts to do phablets.

Windows Phone and Android were not the only alternatives. Nokia was also working on MeeGo, which had a transition plan for their huge Symbian installed base, but was mired in mismanagement. Even after Elop torched his platforms, a small group was working on Meltemi Linux, until Elop noticed and fired them.

With Windows Phone, Nokia was the biggest fish in a small pond. They've managed to sell many phones, primarily by being the only Windows Phone OEM willing to lose money on every device. Microsoft was not "devices and services" back then.

It's good for Nokia's shareholders right now to have a profitable company again, but I'd be surprised if that were the plan. It would have been immensely better to have a profitable company that was also the captain of the industry, and not a 10% market share also-ran.

Hex codes are a good thing

The aversion to hex codes is confounding.

Any competent computer scientist learns hex code. If you don't understand hex, then you shouldn't be holding technical opinions. And average people can't understand normal IP addresses anyway; as far as they're concerned, the dotted quads are hieroglyphs. IPv4 just has shorter hieroglyphic names than IPv6 does.

I find hex codes to be much easier to work with. Each character stands for a unique 4 bits of address. Most allocations are done along half-octet boundaries (prefixes divisible by 4: /32, /40, /48, /56, /60, /64) so each character in the prefix is the same for every host in the network, except for the trailing zeroes in the prefix. Contrast that with IPv4's decimal addresses, where each decimal digit covers several binary digits partially. And IPv4's paucity of addresses means subnets get allocated on awkward bit boundaries.

Concrete example time. Let's say you get allocated 2001:db8:abcd:ef00::/56. Every host on your network will have 2001:db8:abcd:ef00: at the beginning of the address, only varying in the last 16 hex digits, because each subnet is recommended to use 64 bits. If you have more subnets, then the two zeroes at the end of the prefix will change to the subnet address, but otherwise they will all have the same prefix. With the recommended allocation, you have 256 subnets to play with; or you could manually use those 72 bits however you want.

Let's contrast this with IPv4, an allocation of 172.16.64.0/21. Some hosts could have 172.16.65 at the beginning of the address, and others could have 172.16.70, but none will have 172.16.72. Not to mention network masks for hosts that still use those: If you want the final 11 bits to be host address, the mask will be 255.255.248.0, but if you want 10 bits for host address, the mask is 255.255.252.0. You need to do decimal to binary conversions whenever you work with IPv4 addresses. And you have far fewer subnets to play with, or far fewer hosts per subnet.

Hex digits are way easier to use. The vast address space of IPv6 makes it even easier to use. It's not the complexity of the technology that's holding it back, but laziness.

Re: More issues with OpenSSL

I'm going with the theory that the OpenSSL core developers don't deserve more volunteers. SSL is incredibly difficult to do right, and OpenSSL is badly written and badly maintained code. Fixing it is like throwing good money after bad. It should be replaced. The trouble is that it's hard to find another library to standardize on, and to be sure that it's correct.

OpenSSL is written by monkeys.

OpenSSL has exploit mitigation countermeasures to make sure it's exploitable

The discussions also include anecdotes about how hard it is sometimes to get improvements into OpenSSL. But if OpenSSL improves drastically due to Google's involvement, then it may become a good idea to standardize on OpenSSL again.

Re: For sufficiently small values of "wide"

I require a good reason why I would bother learning and using Cobol, when other languages are readily available and widely supported.

With C, I have free-of-cost access to Visual Studio or Xcode or some distribution of GCC or LLVM. With C++, I also have macro systems like Boost or Qt. With Perl, there's CPAN. Python is "batteries included," and there are a bunch of useful libraries. PHP has lots of frameworks. Even Javascript has NPM or jQuery or something. PostgreSQL has very good documentation. I can become productive in almost every modern language without taking special classes or paying for some proprietary toolchain.

It doesn't feel like this is the case with Cobol. GNU Cobol is obviously making things better, but it's implemented in C. I wonder just what is the advantage of using Cobol, when there are so many languages that are better-integrated with my operating systems, and when I personally don't have any Cobol legacy code.