Hacker News new | comments | ask | show | jobs | submit login
Steam'd Penguins (valvesoftware.com)
333 points by mwilcox on July 17, 2012 | hide | past | web | favorite | 162 comments

What will make this truly interesting is if 3rd party game developers follow in Valve's footsteps and start porting their games as well. Because that will hopefully then lead to more attention from graphics card driver makers, solving one of Linux's last remaining weaknesses compared to Windows.

With Microsoft following Apple into the post-PC era (or "PC Plus", as MS COO Kevin Turner likes to call it), Linux will be the last major OS with a traditional desktop metaphor as its dominant UI. Despite all the statements that the PC is dead, I think it will still have widespread uses, at least until brain-computer interfaces become a reality, because touchscreens simply cannot replace keyboards/mice in a large number of contexts.

The Unity Game Engine (unity3d.com) which is very popular will also feature a Linux deployment option from Version 4 on out! Seems like theres alot of movement in the Linux gaming space at the moment, which of course is superb.

Even on Linux, Ubuntu is moving toward Unity and Gnome3.

I like the phrase "PC Plus." It's exactly my experience. I use my desktop and laptop for most things, but for consuming content when a small form factor works my tablet is great.

I think of Linux as more of a black-box (runs your router, embedded) or server (headless) UI. I've never referred to Linux's desktop as it's dominant UI.

Wait, OSX is loosing its desktop metaphor? Did I miss something? And since when is the desktop metaphor essential to mouse / keyboard control?

Having used Mountain Lion, I'd argue its more bringing mobile innovations and things that are successful there to the desktop. To be honest the desktop hasn't fundamentally changed much for a long time so its nice to see something new and being tried. That said, I'm not the biggest of fans.

I know it doesn't work well for multi-displays, but I freakin' love Command Center on my laptop when I'm not near an auxiliary monitor... so they got something a little right at least.

It is nice, and I'll admit, full screen applications with command center are quite nice. Its great having a full screen terminal with tmux running, chrome, xcode, etc... I agree its one thing done right.

And then there is Finder... but at least I have the shell, if only I could keep Finder dead or at least not visually running I'd be happier.

From a brief stint using Xmonad, I just wish there was a way to switch between full screened apps without the animation.. If you assign a key and switch rapidly between say, a browser and a terminal, it can get a bit much sliding back and forth..

>at least until brain-computer interfaces become a reality

A little late to the party.

Sent from my iBrain.

While I applaud what they're doing, it's a bit weird how that post keeps throwing around open-source rhetoric when neither Steam nor any of Valve's games are open source. The only regard in which this has anything to do with open source software at all is that the platform happens to be predominately GPL-licensed, and may conceivably become an indeterminate amount more popular if more games than L4D2 get ported.

Is there something unclear about this?

"Our mission is to strengthen the gaming scene on Linux, both for players and developers. This includes Linux ports of Steam and Valve games, as well as partner games. We are also investigating open source initiatives that could benefit the community and game developers."

I'm sure as they work on the Steam client and as they port L4D2 to Linux, they'll start finding and fixing issues in the kernel/OS/related projects, thus contributing to opensource.

they already found at least one bug in the kernel: https://lkml.org/lkml/2012/6/22/338

What 'open-source rhetoric'? It isn't a Valve Open-source blog, it's a Valve linux blog - open source is not really relevant at all to this post.

It's mentioned three times in the article, including: "We want this to be a community of open source game developers, communicating with each other and talking about current efforts and future efforts in a powerfully creative environment."

And the most damning:

>After all, isn’t that what open source is all about – the idea that collaboration and teamwork achieve amazing things?

Damning in the sense that a game company wishes to sell games commercially? God forbid that they do such a thing!

Yes, I suppose it is better for Linux if nobody attempts to sell commercial software which could work with Linux, ever.

Don't be melodramatic. Damning in the sense that open source, the idea, is not what they are interested in but are riding it's coattails in order to encourage developers who are interested.

This thread is a little overblown and it's too early to point fingers and thump chests - lets see what they put on offer first.

Unfortunately it's never too early for the Linux union to go big on the blame and drama. Sometimes the product doesn't even have to be out before someone starts the bile because they can't see the source.

It is ridiculous to begin talking about things as "damning" when a company is doing nothing but dedicating resources to the cultivation of Linux as a platform for video game development and (yes) the sale and consumption of commercial games.

The word "damning" should not even arise. Don't lecture me about thumping chests and waiting to see what they put on offer, I am responding to inflamed language being applied to a company for no reason other than that they sell software.

Relax, it was just bad word choice.

I meant damning in the 'Valve has been accused of miss-using the term open source' and here is perfect evidence, sense.

I love what Valve is doing and I want to encourage them. Still, their usage is just a bit, 'missing the point'.

He means that a closed source platform isn't exactly conventional open source collaboration. Sure, it might be a boon anyway, but that doesn't make the rhetoric accurate.

It is, since they recently hired SDL developer - and it appears he will be working on it full time, same goes for darkplaces engine creator. So this is direct help to open source libraries.

That is awesome news. Always glad to hear positive bits on the SDL front. http://lists.libsdl.org/pipermail/sdl-libsdl.org/2012-July/0...

They've been hired to work on their respective libraries and publish the code openly? What's Valve getting out of that?

these guys know linux graphics and audio stacks from the inside out. having such people around the office which you can come to, ask pretty much any question and get a meaningful answer is practically priceless IMHO.

Their libraries (especially SDL) are used when games get ported.

It seems like Microsoft has really tied Valve's hands here. Unless something has changed recently, all apps installed on Windows RT will have to come from MS's app store (marketplace? bazaar?), which doesn't bode well for the future of Steam on the platform. I'm hoping they end up doing a lot more than porting Steam to Linux, because outside of Valve's efforts, most of the games on there are not (and will likely never be) Linux compatible.

This would mean that (pretty much) the only way you could game on non-Apple hardware would be to install Linux, which would make Linux the go-to OS for gaming, something I doubt Microsoft would let happen.

I really doubt Microsoft wouldn't allow Steam to run on Windows.

They will still allow people to game, just you will have to buy all of your games via the Microsoft Store rather than Steam/Normal Download/DVD

So, there are multiple approaches to Linux software distribution (for free-as-in-beer closed-source software, like the Minecraft client, Skype, Chrome or Steam):

1. Pump out versioned tarballs, and let distro maintainers deal with packaging. 2. Create a very thin launcher, and deal with updates yourself in the Launcher. Minecraft does this. 3. Provide package repositories for Ubuntu and Fedora and deal with updates yourself. Chrome does this.

I think (1) and (2) are preferable. (1) is also the easiest to implement (!), but may result in delayed distribution to the end user.

It is possible to mix (1) and (3), of course.

The chromium project releases their updates for Ubuntu users thru a PPA, but Canonical also distributes stable packages. A mix that is working really well.

God forbid #2, a system-wide system for downloading and updating applications is one of the zen-things about using the Linux-based OS's.

How it works on Mac OS X and is chaotical, and the way Microsoft does it, with countless messy installers and self-updating apps is hell to deal with.

Well, Steam is by necessity an independent package manager. So you can't really have that Zen if you want to use it. Might as well update itself, too. ;)

To pinkbucket: you seem to have been hell-banned when your account was new. You will need a new account. My condolences.

(3) is much easier to implement for Ubuntu due to PPAs. It's also the most convenient for the users, although trusting a wide range of PPAs is somewhat of a security issue.

It's also an usability nightmare, as a lot of things that Steam is used for both update very often (weekly to daily), and always need to be kept updated to the newest version or they fail to start. You'd have to run a apt-get update && apt-get upgrade every time you launch your favourite game.

Most likely, Valve will provide a steam_installer.deb, which depends on all platform libraries Steam needs, and which only contains a script that creates the necessary users and rights, reserves some directory for steam, fetches the actual Steam off a Valve site, authenticates it, and runs it.

Steam has it's own usability issues though. For example it won't let me play a game and download a game at the same time.

I guess this is for performance reasons, but I'm sure my PC can run Braid and do some disk IO at the same time.

I also can't find a way to have different games installed on different partitions/drives. This would be very useful if you have an SSD but can't fit all of your games onto it.

> I also can't find a way to have different games installed on different partitions/drives. This would be very useful if you have an SSD but can't fit all of your games onto it.

You can do this by moving the directories in question into the other disk, and then creating a junction (symbolic link) from the original location to the new one. On windows 7: mklink /j "path/to/steam/steamapps/common/gamefolder" "path/to/the/other/disk/gamefolder"

Thanks, will try that.

Shame this isn't made a bit easier.

>Steam has it's own usability issues though. For example it won't let me play a game and download a game at the same time.

You can do it manually. Activate the game to play, alt-tab to Steam. Library, view the downloads. The download will be suspended, pause it and then you can unpause it and it'll resume downloading. Alt-tab back to the game.

Lots of work, but it works (or at least I think it did when I last tried it).

The post I was replying to was talking about Linux software distribution in general. I wasn't suggesting using the package as a delivery mechanism for Steam software, which seems like a pretty silly idea.

For Ubuntu, you can release closed-source software in the Ubuntu Software Centre.

Also PPA-based, by the way.

Honestly, I dislike the way Chrome works on Ubuntu. I don't like using PPAs and I'm never sure how/when Chrome gets updated. Sometimes it seems like the OS does it and sometimes it seems the Chrome updates itself (just like on other platforms). Maybe its just me.

So, I think my vote for how Valve updates would be (2). Give the Linux vendors something to package but since they move very slowly you need to be able to fix bugs in a timely manor.

In my experience with Ubuntu, Chrome gets updated through Update Manager like everything else.

Among other things, porting their software to Linux might bring other fruits:

* You can run many instances of a dedicated game server (roughly in the hundreths mb of ram usage) on the same OS, or through virtual machine so it could be moved.

* Gaikai/OnLive live kind of game streaming - again no cost for OS - so they can run each instance of the game per one installed OS and stream video for the rest.

* Console - probably customized version of Linux, and if possible the game running in ring 0 with the kernel. Drawing almost directly to the hardware, without going to API's, user and then kernel functions. (Avoid vertex/index buffer checks, and many other things).

> You can run many instances of a dedicated game server (roughly in the hundreths mb of ram usage) on the same OS, or through virtual machine so it could be moved.

As this very blog post explains, Valve's dedicated servers have always supported Linux.

> Console - probably customized version of Linux, and if possible the game running in ring 0 with the kernel. Drawing almost directly to the hardware, without going to API's, user and then kernel functions. (Avoid vertex/index buffer checks, and many other things).

A system like that would be so different as to barely benefit at all from the reality of what Valve is doing.

It allows them to amortise porting efforts with sales on multiple platforms.

Games running in ring 0? I don't think even consoles do that anymore.

they do, except they run in a hypervisor. no idea if that counts.

No mention of the fact that a bunch of games on Steam already have Linux versions? Or do their blogs usually ignore non-Valve software?

Yeah it would have been nice to see that mentioned. I was curious at how many games could possibly launch with the Linux Steam launch, and there were more than I realized:


I buy mostly indie games, usually the Humble Bundles, so a very good chunk (maybe most?) of my library of 100ish games have Linux versions.

Not only that, but they are also available on Desura, which supports linux.

My biggest hope out of all of this is that Nvidia will get its act together and release proper drivers for Linux now. I honestly don't care all that much about games except that they might force Nvidia to actually care about Linux.


nVidia has always cared about Linux. They've supported non-Windows platforms well for a very long time. They've done a lot of work to provide a good workstation experience on Linux. Today the notable exception is Optimus, but I'm hoping that's going to come down the pipe as nVidia pushes its inevitable KMS branch further down the line.

AMD drivers are still not really usable; you can work around some of the bugs with open-source drivers, but then your speeds are severely crippled.

The mantra among Linux users has always been "buy nvidia". Those with ATI quickly grow to regret it in my experience; a meaningful illustration may be that my parents have two Linux boxes, both relatively low-end PCs with onboard graphics. One of these is almost four years old now, and has an onboard nvidia GPU. The other was purchased three months ago and has an onboard AMD GPU ("APU" according to marketing material. Bobcat? Or was it Bulldozer?). We still use the 4-year-old machine for gaming there because the nvidia drivers are actually sufficient to play the game; the new machine is using an open-source driver due to Catalyst bugs and runs games at probably 20% the performance of the machine 16x its age. That's the experience you get with nVidia's competitors.

Sounds like you missed the discussion on Torvald's "Fuck You" video and nVidia's response. nVidia only cares about Linux when it's easy.

Linus has his reasons to be pissed. But from a desktop users' point of view, Nvidia has been the only vendor that has shipped working GPU drivers with 3d and video for the past decade. All other vendors have had worse drivers, sometimes even completely unusable or just buggy. The situation has improved, though.

There is no way I'll believe getting good performance out of nvidia cards on Linux was easy.

Torvalds was angry about Tegra, not GPU drivers.

Possibly, but their GPU drivers are also frustrating crap. Yes, AMD/ATI's drivers are far worse, but nvidia is still dropping the ball. Binary blobs in your kernel are bad for everyone. Kernel developers are often rightly uninterested in kernel problems that occur in environments that are "tainted" by black-box code. Nvidia's driver doesn't play nicely with the Linux GPU stack and ecosystem. Consider their "twinview" nonsense, requiring a legacy xorg.conf, and duplicating functionality present in the standardized xrandr 1.2. Fullscreen programs are an iffy proposition at best with nvidia's drivers, made even worse if you have multiple monitors, as a result of their refusal to work with the community and the standard tools.

A game developer as important as Valve officially supporting Linux has the potential to force nvidia to finally start playing nice, or even better, get their head out of their asses and just open-source their drivers and get them into the kernel. That way every nvidia user on Earth can benefit from an officially supported, open, community-driven driver.

Slim chance, but I would've said the same about native L4D2 if you'd asked me 2 months ago.

The nvidia driver codebase is HUUUGE, much of it is valuable proprietary information, and much of it is third party. It would be extremely challenging for nvidia to open source it, from both legal and business perspectives.

I do hold out hope for nvidia gaining full xrandr and kms support at some point. The latest beta drivers have significantly increased xrandr support.

Twinview is an implementation of Xinerama, which predates RandR and was standardized for X11 6.4. When nVidia released twinview, it was using (at the time) the standard way to achieve this functionality. They haven't moved to RandR or KMS, which is a bit of a pain granted, but that's a case of being stuck on older tech, not a case of ignoring what's out there.

Add this to your /etc/environment



Will fix the dual monitor/twinview issue for some programs.

It was mostly about desktop GPU drivers and he said nothing about them specifically. And it was most certainly not targeted at Tegra drivers. Tegra kernel parts are mostly open sourced (not the user space, though) and the Tegra kernel team is doing a lot of work that is important to running Linux on ARM SoC's, not just Tegra specific parts.

There's also the Optimus "We have no plans to support Linux" situation.

Well I will not put money where my mouth has been and buy Left for Dead 2. I have been wanting a lot of Valve games, but I swore off all non-native games after Minecraft ( if Notch can do it so can everyone else )

Minecraft isn't really "native" since it targets the Java VM.

And it still had Linux specific issues. That weren't fixed for over a year (there were community patches though). Also, Minecraft's "graphics engine" (raw opengl?) isn't comparable to Source. In any way...

1) source can't do what minecraft does. 2) what do you think source will use in linux? directx? it's opengl or the highway.

Something that impressed me with minecraft was the audio quality. Generally Java audio is rubbish: very rough at the edges and even unstable. Minecraft is using paulscode and it seems to be solid.

What do you mean by "non-native"?

something that needs WINE to run. I want a Linux binary I can run.

Did you miss this part: it now runs natively on Ubuntu 12.04.

Right which is why I am buying it. They have earned their money

Oh! You said "I will not put money where my mouth has been" did you mean "now"? :)

Well thats one really bad spelling mistake on my behalf. I can't even change it at this point. I meant now.

Money spent cannot be eaten (or.. put in one's mouth)

It doesn't mean eating, it's about talking. http://idioms.thefreedictionary.com/put+money+where+mouth+is

Haha, glad you put it together, I had one confused look on my face reading up and down that thread repeatedly.

He probably means by running the games using something like WINE.

I'm going to show my lack of linux history here... why not Debian? Are they really building something that is native to ubuntu that you wouldn't work in <insert-debian-fork crunchbang or mint>?

They're working with Canonical to make sure that it's well supported, and they chose them because Ubuntu in particular is focusing on making Linux for the mainstream. In this context, "supported" on a particular platform doesn't mean "can run," it means "should work out of the box for absolutely everyone." I have no doubt that, come the release on Ubuntu, unofficial packages will come into being for nearly every distro--and on Debian you probably won't even need that.

Ubuntu's the only major distribution (that I know of) that has an integrated and supported process for installing the closed video drivers you'll need to run their games on ati/nvidia.

Arch linux isn't all that hard, not sure if this counts as "integrated and supported"

But... # pacman -S nvidia nvidia-utils


Is all I did for the work install, took longer to finish the setup but thats more due to me having 2 monitors than anything else, one is mad simple. Can't speak to the ati stuff. I know SuSe is about as simple as well.

> Arch linux isn't all that hard

Yes. Yes it is.

(I wrote all this then realized I've gotten off the point, this isn't really addressed at your comment. I recognize that the typical computer user couldn't give a damn about customizing their system and certainly wouldn't bother with setting up Arch, which makes it a poor gaming platform. I'm just tired of constantly hearing about how hard Arch is. It isn't hard, it just takes a bit of effort).

I'd say Arch is pretty simple, just not easy [0]. Easy would be inserting the install cd, walking away, coming back and being able to hit the internet. Which is fine, if that's all you want to do. Arch installation isn't easy, it's simple, and more importantly it works. When I install windows from scratch you know what it comes with? Nothing. No drivers, no programs, nothing. Not even an ethernet driver for me to get the other drivers. Arch comes with all of it out of the box, you just need to go through the tutorial so you can choose which things in the box you actually want.

Arch is simple in that if I go to any scary looking file in /etc and open it up I won't be completely baffled by what's in it. I can randomly shut down any service running on my machine and I know my box will keep chugging and I can recover from it. There's no weird services that I don't know what they do, no scripts that run on startup that I can't understand let alone find. There aren't inter-dependencies that I don't know about. When I run ps aux I can account for every single thing running.

And even then, the package management is simple AND easy. I have never had a problem installing something with pacman, not once. There's this oft-repeated line about upgrades breaking systems, but I've never seen it. If an upgrade fails I look at the front page of archlinux.org and see that some manual commands are needed, copy-paste them and voila things work. I've used arch on three computers now, two laptops and a very old desktop, and have never had any problems. I know there's going to be some horror stories, but I can counter any of them with a quick search for the same in windows. Problems upgrading happen, but they're not common, and not something that's inherent to Arch like I've seen people suggest.

[0] http://www.infoq.com/presentations/Simple-Made-Easy

I'm a systems developer, not a distro maintainer. When I boot off of that iso, I want a menu that asks me this:

[ ] Desktop

[ ] Server

[ ] Embedded

[ ] Custom

This ultimately comes down to a user interface issue, but it's different than the kind of user interface often associated with the HN crowd. It's more like an API user interface, be that a web or programming library API. A crucial feature of good APIs is 'graceful degradation' in terms of easiness/control. Take a look at the libcURL library[1]. cURL has 3 levels of access: "easy," "share," and "multi." If you don't care about a lot of features and just want to download a webpage, the easy interface will get you what you want in a few lines of code. If you have something more complicated, you can go all the way down to the multi interface to access many of the raw functions of the API.

They key takeaway is giving the user control of her complexity level. If I'm a sysadmin compiling an image meant to be deployed on thousands of servers, then something like Arch would be pretty great -- I can fine-tune things how I want. If I'm just a random developer, then I would like to stick to 'sane defaults,' with the option to revisit those defaults later if needed.

I really don't care what cron I have, don't force me to pick one against my will.

[1] http://curl.haxx.se/libcurl/c/

Every time I've installed Windows, it's come with basic drivers and programs.

Short history lesson here then: Until recently installing Windows (except OEM) meant running driver CDs at least for stuff like network drivers (so you could connect to internet to download the updated drivers).

Short history lesson: I have successfully gotten my Windows installs onto the internet without using driver CDs since at least Windows 98SE.

I dunno what I'm doing that's so magical; I just install the dang thing.

This totally depends on the network card chipset. A fresh Windows 7 install on my PC requires external NIC drivers before I get on the net.

Same here, I installed Windows 7 on a PC with a fairly recent Asus MB and I had to install the drivers from the CD.

Running Arch Linux on a Raspberry Pi, it's pretty awesome as a next step distro beyond Ubuntu, but yeah... as an accessible platform for gamers of all kinds, ah no :)

I'll admit, i'm biased. >.<

But having installed Linux back in 1997 and dealing with X11Rwhatever it was back then and minicom and all that jazz, kinda makes my view of hard a bit off of what most people view as hard.

The problem I ran into with Arch was that it forced you to make many decisions about your system up-front, before you had a system to begin with. I even found Gentoo easier to set up, and Gentoo forced me to pick which version of cron I wanted to install.

One nice thing about Arch though is the ArchWiki: it has copious amounts of documentation on software that is usually applicable to other distros. A good example is https://wiki.archlinux.org/index.php/Awesome

looks like it was a good thing i hedged ;)

I think it has to do with their familiarity with Ubuntu server and the popularity of the distro.

They wrote on that post that they do indeed want to support other distro's in the future. I suspect they just want to get some stability before committing to a wide range of distro's. Ubuntu is very popular, so it makes sense.

Distro's can take the .deb files and extract them through their package managers to fit their systems, probably without much hassle.

If it works for Ubuntu it will work for Mint. Mint is 100% compatible with Ubuntu, you can even install the Ubuntu software center in Mint.

I hope so, because I like Mint a lot more than regular Ubuntu.

I expect it will eventually work on any recent distribution, and there would be community docs everywhere on how to install dependencies, configure and run their software on distribution X.

Saying that, Ubuntu is popular, there may be some commercial connections with Canonical, and picking one distribution to do the testing should streamline their efforts.

Debian has the Debian Free Software Guidelines and the community as whole is less supportive of proprietary software, not to mention that it's all volunteers. I think Valve would rather work with a company-supported distro like Ubuntu, which focuses on just working.

In any case, the client may very well run on Debian, it just won't be officially tested or certified by Valve (for now).

Simple answer is that most of the differentiation between Ubunu and Debian is in the areas of audio and video. Specifically regarding drivers, which is a critically important component of game development.

The Linux Standard Base keeps most of the system structure and ABI consistent across distributions, which means that binaries are trivial to move among them.

This is why I think they just miscalculated. They still dont know the religious wars of Fedora vs Ubuntu, or rpm vs deb.

I was seriously hoping for a Valve distro, which would be perfectly tuned for all things media (they hired the SDL guy). Maybe have Steam as the package manager for everything.

Instead, they just joined a camp - this is not good.

The other way they could have approached it was to support Android - not as a playing system, but rather as a console device connected to peripherals - and sponsored some project to run Android apk on Linux with access to the sound and graphics layer (SDL?? )

Instead, they just joined a camp - this is not good.

This is exactly what they should be doing, not further fragmenting an already gratuitously fragmented market.

Every time some commercial vendor makes a move into Linux and people start bickering about distros I just /facepalm.

actually I am indeed talking about the "bickering" that is going to start. I use Ubuntu myself and am most definitely not bickering about Fedora.

However, the Linux market is already fragmented. And even worse, it is deeply religious. Just google for the bickering that ensued when Meego chose rpm over deb. People have already started talking about why not Arch, etc.

The point that you are missing is that, if anyone could unify the Linux fragmentation it is Valve. Steam is a killer app, and I can totally see people migrating towards a distro built around it (and laptops certified for it).

I want Valve to succeed - how will it succeed in a severely fragmented market ? Can it justify the cost of its Linux team then ? Android is having a similar problems in version fragmentation. People have posted articles here on how they are unable to justify multiple version (distro?) development. Multiply that problem by a thousand - different audio toolkits, poor driver compatibilities. I am not able to see how they are going to succeed, unless they unify atleast the multimedia toolkits (something like CoreAudio)

Steam is a killer app, and I can totally see people migrating towards a distro built around it (and laptops certified for it).


Anybody willing to switch distros just to run a game might as well just install a Windows partition.

Eh, switching from Fedora to Ubuntu would not be a big deal at all. The differences really are quite minimal and the actual install would take what, half an hour tops? The effort to install a windows partition would be much higher.

I think they are responding to the suggestion that Valve releases their own distro.

Would certainly take me significantly more than half an hour to switch to Fedora though since I would need to re-learn the package manager, move all my settings across and figure out the subtle differences in configuration etc.

Valve releasing their own distro would be just craaaaaazy. That would be like Amazon or Google doing it. Why bother? There is an awesome one out there for free (Ubuntu).

If Valve wants to make their own console, which is the more likely case, they can license Ubuntu support for cheaper than they could make a terrible distro and it would be win-win-win (which is, of course, better than a win-win scenario).

Amazon or Google is doing it - ChromeOS, Android, Amazon's fork of Android....

If you keep a seperate /home the process is greatly streamlined. I've had the same one through countless ubuntu , gentoo, and fedora installs. Just have to swap out your / and recreate your user.

The only real differences in package management for I'd say 95% of use cases is if you type 'yum' or 'apt-get' before typing 'install whatever'.

From a gamers perspective, Win/ is great but if Win( comes with all that mobile UI stuff Linux (I don't care about which version exactly) will have quite a chance. Given decent drivers for the hardware and being open but not having to much over head will exactly be what a lot of gamers want, it will be running games.

If someone comes up with a upgradeable open pc-based console of sort running steam, well THIS could be the final nail in the traditional PCs coffin. Actually loving playing games on a desktop I hope that doesn't happen any time soon!

Very true, this reminds me of the failed Corel Linux years back.

If the attitude towards fragmentation is to fork a new distro for every application, am I supposed to restart my computer every time I want to task switch or run hundreds of virtual machines?

There are some specialised situations where a custom distro makes sense, like appliance type applications (smoothwall, backtrack etc)

I'm not sure this is the issue you think it is.

Games in the Humble Indie Bundle releases have Linux builds. (.tar.gz or .bin) I haven't had major issues beyond installing the libraries that I get prompted are missing. That could get automated, especially now PackageKit abstracts away from the details of different package management utilities.

Also, you could cover a huge portion of the linux market by just supporting Ubuntu and Fedora. Other distros (e.g Arch) have a community that has some experience in repackaging these for their distro.

I don't think they'd really reduce fragmentation by creating yet anther distro.

They could probably solve a lot of issues with fragmentation by establishing steam in a chrooted environment on the system so that it could maintain it's own versions of certain libraries etc. This could then be targeted by developers as opposed to targeting a specific distro. In essence, a "distro within a distro".

No kidding. How ungrateful is this reaction (and I'm not normally one to grate-guilt)? If it works in Ubuntu, it will probably work just fine on other Linux variants, but Ubuntu is the primary support scenario.

And why would a new distro make sense? Now you can have gaming if you just install ~~Windows~~ ValveLinux.

They clearly mentioned in their post they would support other distributions if STeam on Ubuntu is successful. They are not choosing sides, simply starting with the largest distribution. Which makes sense.

Porting between Ubuntu and Fedora, or even Ubuntu and Android once you have a native Ubuntu build is far easier than porting from Windows to Ubuntu.

I'm glad they are not making their own distro. If I wanted to run a "gaming OS" I'd use Windows or buy an Xbox.

Where they could provide value though is to create a stable platform cross distro that is well supported for game/app developers. Anything that allows porting to Linux without worrying about distro fragmentation is a good thing.

Besides when it comes to more esoteric distros I'm sure they will be able to tap into the community to help bring it there.

I wonder how much of this is really Valve's effort into making Linux a Gaming platform and how much is just to put pressure on Microsoft to relax the policies on Windows 8 that would make the Steam model invalid.

If this path will lead to a world where gamers don't depend on Windows, that would be awesome for users and really stink for Microsoft.

my personal belief is that valve wants to do a linux/x86/steam game console and this effort is a "let's see what we can do with linux" proof of concept.

Steam and Adobe are the only 2 things stopping me making the switch the Linux.

This. And flakey drivers.

I think this is a big step for Linux on the Desktop. I look forward to the day I can ditch OS X and Windows and replace it with a Unix that has a nice UI, stability, and is open.

Flakey drivers are a really big sticking point now, for certain hardware like video cards (most of my drivers "just work").

I've started using Unity (shock, horror) and the whole thing provides such a polished experience that most of the time I forget that I'm using a desktop linux distro. Problem is that when everything is so well polished, problems stemming from bad drivers (nvidia) become really jarring.

I'm hoping that wayland will go a long way to addressing this somehow.

Wayland can't do anything about drivers. Gallium3D, and manufacturer support for it, might. Most of all, though, if Valve starts supporting Ubuntu as a first-class platform, they might be able to put pressure on GPU manufacturers to improve their drivers.

I thought (correct me if I am wrong), that part of the reason that graphics drivers were so messy was that they had to integrate with X which is archaic and not really designed with hardware acceleration in mind.

IIRC the nvidia Linux drivers actually replace parts of X with their own proprietary stuff which makes things hairy when integrating with software designed around "standard" X.

For example nvidia "twinview" seems to make many applications believe that my two 1920x1080 monitors are actually a single 3840x1080 monitor. I can correct this by using a seperate X session for each monitor but then this stops compiz from working.

While dealing with X is annoying in general, nvidia replacing so much infrastructure was just a case of NIH. Their excuse might be a wish to have as common a code base with Windows as possible.

In fact, none of the GPU vendors with closed-source drivers are providing kernel mode-setting, which would be essential for Wayland.

> For example nvidia "twinview" seems to make many applications believe that my two 1920x1080 monitors are actually a single 3840x1080 monitor. I can correct this by using a seperate X session for each monitor but then this stops compiz from working.

NVidia ships Xrandr in 302 drivers (now in beta, available to ubuntu in ppa:ubuntu-x-swat/x-updates) so this should be a problem of the past.

It's been very annoying though, especially games which try to be clever and position the window centered on the screen or maximize to full screen have been half on the left and half on the right monitor.

I can fix it for many (not all) applications by adding this to /etc/environment.



What surprises me, given the number of programmers that use 2 monitors. How does this stuff not get found when applications are being tested?

Not like it would be a difficult fix either, just test for weird resolutions like 3840x1080 then ask the user "are you using 2 monitors?", if they answer yes then adjust the rendering target.

GNU is Not Unix

If you keep to nVidia, you usually get decent performance and decent drivers. ATI is a very different story, that is true.

nvidia drivers still have plenty of issues, especially if you are trying to use more than one monitor and use stuff like flash and compositing.

Adobe is something that makes me want to switch to Linux. I only use Flash from them though.

Valve may be sealing their fate as the most successful game publishing company for this century.

There's a lot of century to go.

Quite so. The differences between 2000 and 1912 are hard to truely fathom. I don't know how you could possibly attempt to predict at that sort of timescale. Even a decade would be pushing it as far as I am concerned.

I don't think it's too long of a time span considering these guys started in 1911:


Outside the Quake series, I think Left 4 Dead 2 would be the most impressive game ported to Linux to date.

I think that Amnesia has a native Linux support and it is a great game.


More recent than the Quakes, Doom 3 had a Linux build.

Not more recent than quake 4

Ah! Good point, I completely forgot that existed.

And Enemy Territory: Quake Wars, which was actually pretty great. It also had megatextured maps, which were nifty.

neither can be honestly called "recent", though... more recent than quake, while technically true, doesn't really say anything.

Will be interesting to see what happens to games sales/usage figures when they release this.

I'm going to guess that most people who are Linux users but are also interested in games will already have a Windows partition that they use for gaming.

So it might be that any increase in Linux gaming on Steam is exactly matched by a decrease in Windows gaming.

I doubt there are many gamers out there who would love to play AAA proprietary games but are holding out until they are ported to Linux.

Well I did not buy AAA games that I wanted to play that were windows only(like Portal 2[1]), but I would bought them if they are ported to Linux, so I would argue that they will expand their user base(a little). I must admit that in case of some games were to good to miss and I played them on wine(like Witcher 2 - my excuse is that it was DRM free and different enough that it was worth rewarding even when there were no Linux version).

edit: [1] - I mean games with no native Linux version(Portal 2 have also Mac version)

Perhaps, but I imagine the number of people in the entire world like this are measured in the 100s.

I just hope they release tarballs of executables or something, because targetting Ubuntu and Ubuntu alone is quite bad. They might say "We'll support other distributions in the future". Yeah, I know that, you will support Fedora.

If you released tarballs (or even open sourced Steam and all your games {just kidding}), it would work, out of the box, on every single distribution.

I sure also hope they release tarballs. But supporting officially only one distribution makes sense. With proprietary games, they are in the land of consumer rights and consumer support. They can only say "we support X" if the software really runs on X, X being Ubuntu here. If it doesn't run, the consumer has the right to get his money back (instead of getting it back out of goodwill).

If another distro patches a driver and thus L4D2 runs worse or not at all (or something like wayland changes the infrstructure more than we expect and another disto moves first), this would be an issue for them without that official focus on Ubuntu.

Hopefully they won't actively try and prevent people from running it on other distros though, providing they find a way to do so.

If you're a die hard Arch/Gentoo/Slackware user you're probably used to not having much in the way of commercial support anyway.

If it'll be anything like the windows version of Steam, it'll likely manage games in parallel with the package manager, thus about as distro-specific as tarballs.

As long as Steam itself is packaged as .deb, .rpm and tarball, it'll probably work as well as other linux games.

In a worst-case scenario, it would not be all to hard to write a script to extract the content of a debian package into a tarball.

I read the title as "Valve Linux Box" and was a fairly disappointed.

I would think that this is just the first stage towards said box. First, they make their software work on the only OS you can ship for free and then they make their own distro, which they put on their box. If they're smart about it, their distro will be fairly common, with few customizations, so that they can pull in a lot of the open source fixes.

I said this somewhere else in this thread, but making their own distro would be just craaaaaazy. Ubuntu is already free, comes in several varieties and is, frankly, better than anything they could do for the cost.

If they make a game console, which is rather likely it seems, they could just license Ubuntu embedded or ubuntu core (can't remember what it is called) for orders of magnitude less than it would cost them to build their own distro. Anything else just doesn't make sense.

Now, not saying they'll do that. The game industry is notoriously bad with NIH syndrome.

Why not releasing it in the Ubuntu software store instead of making a new one?

It would be the sensible thing to do if they cared about gamers on Linux. But I guess Steam is more about owning and controlling the channels of software sales and distribution.

That's exactly what Steam is about and always been about, what else could it be about?

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact