What Should Valve Do For Linux & Open-Source?

[kwahoo]

At 2008 GDC Valve showed a presentation about Source as multiplatform (Windows/PS3/X360) engine.

Porting Windows->consoles (especially PS3) is a very hard task. Low memory, slow storage, in-order CPU etc.

[KameZero]

Option 1:

Port the game from Windows to Linux. Obviously, the graphic engine would have to be re-written to OpenGL, all the Windows code ripped out, everything run through testing, and tested against multiple kernel/distributions to ensure proper compatability. So figure at least 6 developers at a cost of $50k + benefits, over a year, plus a long term support staff of the same number at the same price for about two years. So figuring about $150k per worker per year of work (benefits, remember?), you come out to ~$900k for development and $1.8 Million for long term support. That equates to $2.7 Million in extra personel costs.

From your calculations here I can only assume that you're fully splitting your codebases between Operating Systems, and hiring substandard developers. Even for a large AAA game, it shouldn't take more than ONE good developer to port it given a years time. Hell, 900k just for the porting part? Just hire Icculus, he'll have it done in 3 months and then you can hire a linux dev to support it thereafter.

You're windows devs shouldn't be writing code that can't be used crossplatform. And most don't because they like being able to port to consoles.

[johnc]

Hire me, I'll port it.

But this is why I think we'd really need a Linux-based Steam console of sorts to really get anywhere. I foresee us being in a similar situation to Mac users... we'll have the Valve games plus a lot of indie-type and smaller games. Many of the bigger names won't bother and so people will be forced to stay on Windows as their one-stop gaming platform.

Which is why Valve jumping headlong into Linux doesn't make sense to me from a business perspective (outside of a console option). They might be doing it for hobbyist / "it seems like an exciting thing to work on" reasons but those aren't ultimately enough to significantly change the gaming landscape on Linux. (Or, at least, not make it any more advanced than what is on Mac today.) But casual and indie gaming is becoming increasingly popular so maybe that's all that really matters.

[Veerappan]

Port the game from Windows to Linux. Obviously, the graphic engine would have to be re-written to OpenGL, all the Windows code ripped out, everything run through testing, and tested against multiple kernel/distributions to ensure proper compatability. So figure at least 6 developers at a cost of $50k + benefits, over a year, plus a long term support staff of the same number at the same price for about two years. So figuring about $150k per worker per year of work (benefits, remember?), you come out to ~$900k for development and $1.8 Million for long term support. That equates to $2.7 Million in extra personel costs.

Figuring the game in question sells for $60, lets assume sold to retailers at a profit of $45 per license sold. So to break even, you would need to sell, at a minimum, 60,000 units above what would normally be purchased. (I can't stress that point enough). And that assumes full price of sale. Taking development time into accounts, users may be expecting a lower price at release, which would increase the number of required units to reach break even.

I think you're over-estimating costs a bit here. First, estimates of overhead costs I've seen put the number at closer to 145% of employee wages, and that's for a production facility paying $10/hour. That overhead will go down as a percentage as employee wages go up (the life/health insurance won't go up, but 401k and maybe disability will). I'd say that it's probably closer to $112k per developer.

I don't know who gives full development-stage staffing to long term support as well. Either you'll cut that number by 1/2, or you'll be getting subscriptions or putting out DLC/expansions to give the developers something to do beyond bug fixes and patches. Any game that has made it through to release and gotten its first patch shouldn't require the same number of developer hours as a game in development... At least that's the way that stuff works here (not game software, but it's still commercial software development).

I'd also like to point out utilities like MojoShader. They can't do the full job of porting your shaders and API calls to OpenGL, but between that, using something like OpenAL or SDL, and other cross-platform APIs, you can reduce the development time that would be required if you decided to scrap the entire renderer and re-write it in OpenGL from scratch.

Porting a AAA title from DX9/10/11 is still going to be expensive, I'm not denying that. But I think you've over-estimated the expenses a bit... Although if you're not talking of a simultaneous Windows/Linux release, you've also overestimated the sales price that can be expected (which you alluded to).

[1c3d0g]

When it comes to porting games to Linux, there are basically two options a company can take. For the purposes of this discussion, I'm going to assume that the company in question lacks any Linux staff to start.

Option 1:

Port the game from Windows to Linux. Obviously, the graphic engine would have to be re-written to OpenGL, all the Windows code ripped out, everything run through testing, and tested against multiple kernel/distributions to ensure proper compatability.
...

None of this is relevant if the developers did their job right the first time. If they wrote the game conforming to the specs of OpenGL (instead of DirectX), you'd instantly eliminate most of the hassle of porting (as OpenGL is multi-platform and does not suffer DirectX's platform tie-in). Professional companies like Epic have done this way back then with UT2k4 (yes, I know it had 2 renderers, but both were equally capable), and it became one of the biggest, most-played (I'll say the best) game of all time. All-in-all, this will be a huge boost for OpenGL as they're already the standard when it comes to professional applications like CAD etc. Now that more games will be rendered using the OpenGL, we as Linux users have nothing to fear, we can only gain more users and better support from larger companies, which will make Linux a viable platform for gaming.

[gamerk2]

From your calculations here I can only assume that you're fully splitting your codebases between Operating Systems, and hiring substandard developers. Even for a large AAA game, it shouldn't take more than ONE good developer to port it given a years time. Hell, 900k just for the porting part? Just hire Icculus, he'll have it done in 3 months and then you can hire a linux dev to support it thereafter.

I make the assumption that a DX to OGL conversion will take significant effort to undertake. You also make the (silly) assumption that most developers have a clue how to code properly.

(A good article that matches my view of how most programmers are when leaving school: http://www.flounder.com/bricks.htm And in my mind, most don't get much better over time.)

You're windows devs shouldn't be writing code that can't be used crossplatform. And most don't because they like being able to port to consoles.

First and formost, I put in the disclaimer "when performance matters". Which in games, it typically does. Secondly, you make the assumption there is an intent to go multiplatform, which there typically isn't.

[gamerk2]

None of this is relevant if the developers did their job right the first time. If they wrote the game conforming to the specs of OpenGL (instead of DirectX), you'd instantly eliminate most of the hassle of porting (as OpenGL is multi-platform and does not suffer DirectX's platform tie-in). Professional companies like Epic have done this way back then with UT2k4 (yes, I know it had 2 renderers, but both were equally capable), and it became one of the biggest, most-played (I'll say the best) game of all time. All-in-all, this will be a huge boost for OpenGL as they're already the standard when it comes to professional applications like CAD etc. Now that more games will be rendered using the OpenGL, we as Linux users have nothing to fear, we can only gain more users and better support from larger companies, which will make Linux a viable platform for gaming.

The issue with OpenGL is the API isn't nearly as friendly as DX at this stage. Used it a handful of times, and while it has relative feature parity with DX, ease of use, and the fact that most middleware engines are coded in DX (Unreal, again) means thats the way the majority of the market will go.
There will always be engines that use OGL, and titles using those engines should be SIGNIFICANTLY easier to port over.

My primary point is this: Because WINE exists, why should developers spend time (money) developing a native linux release, especially in the case where the title already runs on WINE?

[benmoran]

Because people won't buy it?

[bug!]

nice picture http://www.ubuntuvibes.com/2012/07/z...ft4dead-2.html