-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
On 23/04/16 23:05, Timothy Pearson wrote:
On 04/23/2016 03:34 AM, Daniel Pocock wrote:
There are also various other ways to approach this, for example, I started a discussion about ARM-based NAS devices on the debian-arm list[1]. This is one market where the hardware is readily available and the fact it is low power is considered a virtue by most purchasers.
The other thing I would encourage is MIPS-based platforms for routing devices. There are a number of options on the market already, some with quite a number of GbE network ports built in.
The ASUS C201 appears very weak in the specs. Some things that bother me are the screen resolution (only 768 pixels high) and it is USB 2 only. I don't like the Chrome logo on it either (or is that a sticker that comes off?). Are there slightly stronger alternatives?
It is fairly weak; there hasn't been a lot of push to create stronger alternatives that still respect the owner's rights. The POWER systems are the only real option on the high end, leaving middle-ground users (including higher-end laptop users) without any options. This probably won't change for a while due to razor-thin margins in the mid-range arena and lack of overall demand.
That all being said, there are some stronger ARM Chromebooks coming down the line this year. Google Oak is one platform to watch; the SoC used in those machines gives you ARM64 plus virtualization, but sadly that machine is not yet available to the public.
Another strategic topic on this theme: people won't necessarily see this thread and throw away all their x86 equipment the same day. However, how can these ideas be introduced to people at the times when they are making purchasing decisions?
You're correct, and we're not advocating that people dump their existing machines, only that this is considered when purchasing another machine. What I personally would like to see is the major FOSS projects and distros starting to think beyond x86, perhaps by compiling and testing their software on ARM / POWER as well as x86, even if they are continuing to develop on older x86 machines for the time being. In the case of router distributions (pfSense comes to mind) the developers should start focusing on providing (at minimum) an ARM or MIPS port; this would also have the bonus of lowering the power consumption of the router package.
Debian already does that:
https://buildd.debian.org/ (see the list of CPU architectures across the top of the table)
Here is an example for one of my own packages built on all architectures. The build also runs the unit tests on every architecture, to catch any big endian issues, etc:
https://buildd.debian.org/status/package.php?p=resiprocate&suite=sid
Another thing that comes to mind: getting some of this hardware out in public at events. What would it take to have some POWER-based systems on display at LinuxWochen and MiniDebConf Vienna (28 April - 1 May) this week, or at DebConf in July?
Vienna would be good because there is a lot of official interest in reducing the dependency on US technologies: http://techrights.org/2009/12/11/osor-updates/
and there are various projects going on there already.
As you say, the current Opteron 4xxx/6xxx series devices are currently adequate for most work. The danger is complacency; 5 - 10 years from now those chips will be wholly inadequate for many common tasks and there may be no way to upgrade in performance without also sacrificing freedom.
I should also point out that all recent AMD chips, including the upcoming Zen Opteron CPUs and the modern FX-series devices, require signed binary blobs to boot. They are not free and, worse, by design, they can never be freed. AMD has also refused to release any documentation on the A1200 series devices, and it is highly likely they will also require a signed TrustZone binary to boot.