Beyond the Code
or, how to win friends, influence people and make a living by writing open source software. It's not just about the code.
Follow me on Twitter as @jonobennett.
Friday 17 July 2009, 5:54 PM
When is a new operating system not a new operating system?
Reading Steve Ballmer's comments about Google Chrome OS, I'm struck by how odd some of his ideas are. Like what constitutes a brand new operating system, for instance. He also appears to think One Size Fits All is a good way to develop software.
Let me set you straight on a few things, Steve. First, Chrome OS is anything but a new operating system. Google is planning to use the Linux kernel, which has been around for quite a while now. The Chrome browser may well be very new, but it hasn't been written just for Chrome OS. Since Google is planning to release the whole project as open source anyway, they're going to reuse as many bits of existing software as possible. It will also share a huge amount of code with Android. You'll probably be able to count the lines of brand new code in Chrome OS in an 8-bit buffer.
So if you're confused as to why Google would want to have two client operating systems, Steve, it's because you can't see that they won't. They're just different packaging of very simliar codebases. Just exactly what are Windows Vista and Windows Mobile if not two client OSs, anyway?
Which brings me to my next point: that the idea of a single operating system to suit any type of client device is ridiculous. It may seem like I'm contradicting myself here somewhat, but that depends on what you define as an "operating system". If it's the totality of software that is installed on a device by its manufacturer — which is close to Microsoft's idea of what an OS is — then each Linux distribution or version of Vista constitutes a different operating system, despite having the majority of their respective codebases in common. If it's binary compatibility, the same codebase on different processor architectures becomes separate OSs. Higher level API compatibility? That's a whole other can of worms. Whichever way you look at it, there doesn't seem to be any definition of an operating system that means you can have the "same" one on different classes of device.
What matters most is having the right mix of software, all the way up the stack from kernel to applications, that suits your device's intended use. Whether that means you have one, two or three hundred "operating systems" is besides the point.
Let me set you straight on a few things, Steve. First, Chrome OS is anything but a new operating system. Google is planning to use the Linux kernel, which has been around for quite a while now. The Chrome browser may well be very new, but it hasn't been written just for Chrome OS. Since Google is planning to release the whole project as open source anyway, they're going to reuse as many bits of existing software as possible. It will also share a huge amount of code with Android. You'll probably be able to count the lines of brand new code in Chrome OS in an 8-bit buffer.
So if you're confused as to why Google would want to have two client operating systems, Steve, it's because you can't see that they won't. They're just different packaging of very simliar codebases. Just exactly what are Windows Vista and Windows Mobile if not two client OSs, anyway?
Which brings me to my next point: that the idea of a single operating system to suit any type of client device is ridiculous. It may seem like I'm contradicting myself here somewhat, but that depends on what you define as an "operating system". If it's the totality of software that is installed on a device by its manufacturer — which is close to Microsoft's idea of what an OS is — then each Linux distribution or version of Vista constitutes a different operating system, despite having the majority of their respective codebases in common. If it's binary compatibility, the same codebase on different processor architectures becomes separate OSs. Higher level API compatibility? That's a whole other can of worms. Whichever way you look at it, there doesn't seem to be any definition of an operating system that means you can have the "same" one on different classes of device.
What matters most is having the right mix of software, all the way up the stack from kernel to applications, that suits your device's intended use. Whether that means you have one, two or three hundred "operating systems" is besides the point.
Thursday 16 July 2009, 1:03 PM
Aussie ICT researchers open up
NICTA, Australia's ICT research organisation, has released a bunch of projects as open source and collected them in one portal called OpenNICTA. NICTA is a government-funded body which aims to be a centre of excellence for computing and IT research. It also aims to commercialise the products of the research it does.
The projects collected on OpenNICTA are quite varied, but all lower level frameworks and libraries, rather than finished applications. Most are released under the GPL or BSD licences.
If the aim of all this work it to maximise the benefit to Australia, why then give away the software developed at NICTA? First, you're giving Australian taxpayers — and don't forget businesses are taxpayers too — easy and equal access to the fruits of the research. Using existing open source licences also means NICTA's administration overhead is virtually nil.
It also means those of us outside Australia get a look-in too, and that the Aussies can get the benefit of developers outside their country who choose to submit patches to the projects.
For publicly-funded software like NICTA's, release as open source is the fairest way of getting value back to the people who've paid for it. Other countries should learn from Australia's example.
The projects collected on OpenNICTA are quite varied, but all lower level frameworks and libraries, rather than finished applications. Most are released under the GPL or BSD licences.
If the aim of all this work it to maximise the benefit to Australia, why then give away the software developed at NICTA? First, you're giving Australian taxpayers — and don't forget businesses are taxpayers too — easy and equal access to the fruits of the research. Using existing open source licences also means NICTA's administration overhead is virtually nil.
It also means those of us outside Australia get a look-in too, and that the Aussies can get the benefit of developers outside their country who choose to submit patches to the projects.
For publicly-funded software like NICTA's, release as open source is the fairest way of getting value back to the people who've paid for it. Other countries should learn from Australia's example.
Friday 3 July 2009, 4:22 PM
Software patents, and how to route around them
Microsoft recently used its patents on some aspects of the FAT filesystem to beat TomTom into submission, with the latter ending up having to change its products, pay Microsoft and license some of its own patents to it. It wasn't a good result for TomTom, even though there were signs of Microsoft being careful in its approach to the GPL.
The other sad thing about the settlement was that the validity of Microsoft's patents were never tested in court. The ideal end to the software patent wars would be for most of them to be thrown out, but that will take time and resources. Patent pools are a good way of coming together to stand up to bullies, they don't solve the general problem. Until the law intervenes, we need to find technical ways around software patents.
Now Andrew Tridgell, a developer working on Samba has posted a patch that retains most of the functionality of the FAT implementation, while removing the functionality that infringes the patent used against TomTom. The price paid is that short filenames aren't generated when an application writes a file with a long filename. This breaks backward compatibility with applications that can only read 8.3 filenames, but if you're sticking to Free software, that's a very small number indeed.
That a patent is bypassed so easily by so small a technical measure should give you an idea of how much merit the patent really has, but as I've said it needs to be tested in court for the threat to go away. Until then, reactions like this from the Free software community are our best weapons.
The other sad thing about the settlement was that the validity of Microsoft's patents were never tested in court. The ideal end to the software patent wars would be for most of them to be thrown out, but that will take time and resources. Patent pools are a good way of coming together to stand up to bullies, they don't solve the general problem. Until the law intervenes, we need to find technical ways around software patents.
Now Andrew Tridgell, a developer working on Samba has posted a patch that retains most of the functionality of the FAT implementation, while removing the functionality that infringes the patent used against TomTom. The price paid is that short filenames aren't generated when an application writes a file with a long filename. This breaks backward compatibility with applications that can only read 8.3 filenames, but if you're sticking to Free software, that's a very small number indeed.
That a patent is bypassed so easily by so small a technical measure should give you an idea of how much merit the patent really has, but as I've said it needs to be tested in court for the threat to go away. Until then, reactions like this from the Free software community are our best weapons.
Wednesday 1 July 2009, 5:13 PM
NYC launches municipal data mash-up contest
Michael Bloomberg, mayor of New York, has announced NYC Big Apps, an annual contest to produce applications using data from the City to increase access and transparency. The City intends to release a number of data sets covering various services, and award a cash prize to the best application written to use them.
There's no mention in the press release of what format the data will be made available in, nor which licence it will be under. While much of the work done by Federal government in the US is released into the public domain, I don't know whether the same applies to local government like NYC.
The contest itself is a good idea, but government data shouldn't just be released in one-off chunks. A truly modern government, whether local or national, should aims to have data feeds available of anything that might be of use to its citizens.
Given a decent API and a licence that allows free redistribution, people will come up with applications that make government more effective, possibly lower costs and almost certainly restore people's faith in being able to get things changed.
There's no mention in the press release of what format the data will be made available in, nor which licence it will be under. While much of the work done by Federal government in the US is released into the public domain, I don't know whether the same applies to local government like NYC.
The contest itself is a good idea, but government data shouldn't just be released in one-off chunks. A truly modern government, whether local or national, should aims to have data feeds available of anything that might be of use to its citizens.
Given a decent API and a licence that allows free redistribution, people will come up with applications that make government more effective, possibly lower costs and almost certainly restore people's faith in being able to get things changed.
