Comment by nl | Hacker News Reader

Comment by nl | Hacker News Reader

Parent Full thread

I think you are completely underestimating (or misunderstanding) what this is.

The Open Compute designs (which is where this will end up) aren't just a list of parts. They include things like blueprints for custom chassis, circuit boards for power supplies and even designs for solid state storage devices[1][2][3]. The also include the source code for custom management software[4].

While it is true that they don't generally design their own logic boards, it is quite a long way from a list of parts.

I don't think backblaze called their case designs open source hardware.

You'd be wrong about that: This new Storage Pod performs four times faster, is simpler to assemble, and delivers our lowest cost per gigabyte of data storage yet. And, once again, it’s open source.[5]

I think "open source" is completely appropriate here, and to be honest I don't quite understand your objection.

[1] http://www.opencompute.org/wiki/Server/SpecsAndDesigns

[2] http://www.opencompute.org/projects/chassis/

[3] http://www.opencompute.org/projects/power-supply/

[4] http://www.opencompute.org/projects/hardware-management/

[5] https://www.backblaze.com/blog/backblaze-storage-pod-4/

I wasn't saying it shouldn't be called "open source". I was saying it shouldn't be called open source "hardware design".

Hardware design isn't limited to gates, PCB traces and ASICs. This seems remarkably like hardware design to me.

As much as you might not agree, plans for building a birdhouse constitute hardware design too.

Plans for building a birdhouse are completely incomparable to what the parent is talking about. This would be like me having an "open source" birdhouse design where the birdhouse is made of glued-together Intel™ processors.

Sure I've opened the schematics, but unlike a birdhouse made from wood you have zero chance of being able to build one from scratch on your own, or sourcing the Intel™ processors from another vendor.

This is the same reason for why if I release the binary of a program produced with a secret compiler as "open source" I'll be laughed at, it completely goes against the spirit of the thing, which is that anyone should be able to change any part of the thing I've released.

It's good that they're doing this. But it's completely incomparable to actual fully open source hardware.

Well IMO "open source hardware design" can have items in their bill-of-materials whose design themselves are not open source. I think that's where we'd end up disagreeing on terminology.

Open source gives you the ability to fork the design. An Open Source birdhouse made of Intel processors could be forked and made from AMD ones.

Sure I've opened the schematics, but unlike a birdhouse made from wood you have zero chance of being able to build one from scratch on your own, or sourcing the Intel™ processors from another vendor.

This is some kind of confluence of the values of FSF captial-F "Free" style software with the more pragmatic goals associated with open source.

We've eclipsed the usefulness of this birdhouse analogy, but suffice to say being able to replace one proprietary commercial vendor with another doesn't make something open source.

    > This is some kind of confluence of the values of
    > FSF captial-F "Free" style software with the more
    > pragmatic goals associated with open source.

Well, pragmatically, if you needed to alter the hardware of this "open source" hardware what percentage of the total electronic complexity would you be working with? I'm going to aim for a conservative one-digit percentile.

There's a real and unambiguous line to be drawn in the sand here. If you get "open source" plans for hardware that you can't replicate with third-party fabrication & your own materials it's not open source. It's just a recipe for assembling commercial proprietary components.

It's awesome that people publish those recipes, but don't call it open source hardware and dilute the meaning of hardware that is genuinely open.

Don't conflate 'electonic' with being the only valuable thing here.

Some of the most active areas of research in data center design are around the mechanical design to maximize cooling.

Couldn't one use the same reasoning to argue that open-sourcing an iOS app isn't really "open source", since you need proprietary hardware to properly run it?

I think the crucial difference is that when you talk about "open source X" it means that any aspect of X can be modified as you please. Obviously when you're talking about an open source iOS app it's understood that the underlying OS isn't open source.

Since there is hardware that is actually open in that sense it's worthwhile to distinguish whether you're talking about open hardware or merely open plans for assembly of closed hardware.

Why not?

Wikipedia defines Computer Hardware as:

Computer hardware (usually simply called hardware when a computing context is concerned) is the collection of physical elements that constitutes a computer system. Computer hardware is the physical parts or components of a computer, such as the monitor, mouse, keyboard, computer data storage, hard disk drive (HDD), graphic cards, sound cards, memory, motherboard, and so on, all of which are physical objects that are tangible.[1] In contrast, software is instructions that can be stored and run by hardware.

It's difficult for me to see how this doesn't fit.

The point is, this does not include the design for any of those components. It would be open source hardware design if it included the design of a monitor, mouse, keyboard, computer data storage, hard disk drive (HDD), graphic card, sound card, memory, motherboard. Open source design of any of those components would be open source hardware design. As it is, what they provided uses closed source hardware components. The hardware parts? They're nvidia cards and intel processors -- not open source. In this case it is the build specification that is open source. I guess there is a blurred line between the two. I would call circuit board designs that include a closed source chip or two as open source hardware. I just think the line should be lower level for it to be called "hardware design" as opposed to "build specification".

Reply on news.ycombinator.com