Linux Foundation: Bugs Can Be Made Shallow With Proper Funding

jones_supa writes The record amount of security challenges in 2014 undermined the confidence many had in high quality of open source software. Jim Zemlin, executive director of the Linux Foundation, addressed the issue head-on during last week's Linux Collaboration Summit. Zemlin quoted the oft-repeated Linus' law, which states that given enough eyes, all bugs are shallow. "In these cases the eyeballs weren't really looking", Zemlin said. "Modern software security is hard because modern software is very complex," he continued. Such complexity requires dedicated engineers, and thus the solution is to fund projects that need help. To date, the foundation's Core Infrastructure Initiative has helped out the NTP, OpenSSL and GnuPG projects, with more likely to come. The second key initiative is the Core Infrastructure Census, which aims to find the next Heartbleed before it occurs. The census is looking to find underfunded projects and those that may not have enough eyeballs looking at the code today."

Comments

Linux was better when there was little funding.

[Anonymous Coward]

I've been using Linux for an awfully long time, since the mid 1990s (Yggdrasil, then Debian). Over time, as Linux has gotten more and funding, it has gotten worse and worse. I initially switched to Linux because it generally just worked, and it worked better than many of the alternatives. But now it's just getting fucking horrible. I mean, look at systemd. Normal users, and especially power users, don't want it. It just causes problem after problem for many people. Yet we have corporate interests and corporate-funded developers forcing it on us, even forcing it into community-oriented distros like Debian. GNOME and Firefox are other great examples of community-based open source projects that got co-opted by money and ruined, to the most recent versions of both being almost totally unusable. On the other hand, we see projects that get less commercial interest, like Slackware and Xfce, producing the most usable and reliable open source software systems around. Linux was better when there wasn't so much money floating around. Back then it was about creating great software, and doing things right. Now it's about everything but that.

Re:Linux was better when there was little funding.

[BarbaraHudson]

As Heinlein noted, TANSTAAFL, just like there's no such thing as free beer. Everything has a cost. Even free software.

And when you have such a fragmented ecosystem, the attack surface is going to be huge (after all, an OS is more than just a kernel), and the idea that "with enough eyes all bugs are shallow" is patently false. So it turns out that open source has been to a large extent relying on the same "security through obscurity" model. This was fine a decade ago, but the competition have stepped up their game and can afford to throw money and bodies at the job without begging.

The solution would be to do a code freeze for 2-3 years while the developers of the various projects audit their code and the ways other projects interact with their code - not just for security problems, but to get rid of bloat and cruft. That's not going to happen, because it makes too much sense. Everyone wants the newest shiny.

Linux was definitely better when there were fewer distros. What a mess.

Re:Linux was better when there was little funding.

[phantomfive]

Yet we have corporate interests and corporate-funded developers forcing it on us, even forcing it into community-oriented distros like Debian.

Debian adopted systemd for reasons outlined here. It wasn't a conspiracy. Poettering knew that distros are crucial to the adoption of systemD, so he's made things as easy as possible for them, and given them features they wanted. Essentially systemd makes it easier to write an init script, and since Debian writes a lot of them, they liked that.

Of course, Poettering has been less responsive to other parties (like the kernel devs), but that's another topic.

Re:Linux was better when there was little funding.

[BarbaraHudson]

As a programmer, I find systemd horrible. It breaks compatibility with Unix. It's a nightmare that will shrink the open source landscape to just linux. The rest of us must now reinvent basics like toolkits, web browsers, etc because Linux zealots want to take over the world.

Or you could just bang around on this.

Pre-emptive multitasking with 1000hz scheduler, multiprocessor, multithreading, ring-3 protection
Responsive GUI with resolutions up to 1920x1080, 16 million colours
Free-form, transparent and skinnable application windows, drag'n drop
SMP multiprocessor support with currently up to 8 cpus
IDE: Editor/Assembler for applications
USB 2.0 HiSpeed Classes: Storage, Printer, Webcam Video and TV/Radio support
USB 1.1 Keyboard and Mouse support
TCP/IP stack with Loopback & Ethernet drivers
Email/ftp/http/chess clients and ftp/mp3/http servers
Hard real-time data fetch
Fits on a single floppy, boots also from CD and USB drives

Since the blurb was writtten, browser, digital tv, webcam, movies, etc. have been added.

Or you can play around with Plan9 - runs by itself or as an application atop linux, windows, etc.

Re:Linux was better when there was little funding.

[Skiron]

" like Slackware and Xfce, producing the most usable and reliable open source software systems around." Well said - exactly what I use (even Slackware on my Raspberry Pi's).

Re:Linux was better when there was little funding.

[tlhIngan]

Over time, as Linux has gotten more and funding, it has gotten worse and worse. I initially switched to Linux because it generally just worked, and it worked better than many of the alternatives. But now it's just getting fucking horrible. I mean, look at systemd. Normal users, and especially power users, don't want it. It just causes problem after problem for many people.

No, it hasn't gotten worse. It has gotten responsive to user demands.

Back in the 90s when life was simple, users were simple. Unless you used an Amiga or MacOS, if you played a sound, that was it - no one else could play a sound (MacOS and Amiga had software mixers so you could listen to music AND hear application generated sounds - you could use exclusive mode if you needed it, though).

Likewise, you logged in and you rarely had things starting up just for you.

And your networking options were... single. You either had Ethernet, or a modem, and only one IP per host. And rarely did you move - I mean, if you were on Ethernet, it was assumed you were on the same network permanently, or at least changes were rare.

Nowadays, user demands have gone way up. Audio has to be mixed by the OS because the user may listen to tunes, start yakking on VoIP, and having sound effects played while gaming, all simultaneously. The VoIP call goes over say, a Bluetooth headset or the communications path, while the music and sound effects play through the main speakers. Oh, and no application is to dare use the HDMI port to send audio as it's hooked to a monitor with no speakers. A modern PC can easily have 4 or 5 different ways to play audio.

Likewise, when you log in, you probably have a few per-user services you like to have - either from the environment you're using or other services. It would be a shame if logging in again restarted those services (e.g., you log in locally, then log in remotely over ssh) or if those multiple sessions couldn't communicate with each other (e.g., you make a change remotely, and it fails to propagate through the rest of the logins).

And networks... well, an Ethernet port or WiFi? A user may connect to many different networks in a single day, and have more than a few ways to send a packet around. Perhaps they're hooked to their same network multiple ways - either dual Ethernet, or Ethernet plus WiFi. And maybe the next time the connection is re-established, those ports need to be firewalled because it went from private network to public.

Back in the old days, well, audio was simple because your PC couldn't really do multiple things at once. Networks were generally safe so it didn't matter that you didn't bring up the firewall on the public Ethernet connection. And users didn't run too many things in the background because no one could imagine needing to log into the console AND over ssh simultaneously, or they could just remotely kill the session because there wasn't important stuff to save.

And it's perfectly fine on a server that sits in a rack and never moves until it's powered down and retired. But modern users need this complexity just to manage their normal use case. Sure you can force the user to tell you what kind of network is at the other end, or to re-establish the VPN, but users want computers to do stuff automatically - I mean, why should I tell the computer this coffeeshop WiFi is public over and over again - can't it remember?

Or to reconfigure my VoiIP app because I attach my Bluetooth headset to my computer so it now uses that - why can't it ask for a communications headset, and if one isn't available right now, use the default audio hardware. Then when one suddenly appears (Bluetooth!), automagically use that? Zero reconfiguration, event he app doesn't have to reopen the audio device because the audio core did it internally.

It should be telling that the most popular Linux "distribution" in the world is Android, which has its own init system (like systemd, it manages processes, events, and other things), its own audio

Whose Eyes?

[Capt.Albatross]

Even for non-security bugs, the many-eyes hypothesis contains a large dose of wishful thinking, but at least in that case most eyes are looking with the same purpose. When it comes to security, however, it is a race between black-hat and white-hat eyes, and the former only have to win once.

Re:Whose Eyes?

[thieh]

The problem with many-eyes hypothesis is that not everyone looks for the security bugs and not everyone is capable of looking for every kind of bug. Everyone will notice a problem when the UI behaves differently. Not many of us will notice when the command-line utility did something subtly different (especially when it comes to RNG, like the Debian OpenSSL bug back in 2008) without abnormal output.

Re:Whose Eyes?

[swillden]

Even for non-security bugs, the many-eyes hypothesis contains a large dose of wishful thinking

Not true; Torvalds' observation wasn't what he wished would happen, it what what he'd observed repeatedly on a large and complex project over the period of many years.

That said, I think your disagreement is because, like many, you misunderstand the hypothesis. What Torvalds said wasn't that given enough eyes all bugs are visible, but that they're shallow, meaning easy to track down and fix. The hypothesis doesn't even come into play until the existence of the bug is known.

And, undoubtedly, there are some bugs that are deep no matter how many eyes are looking at it, so the hypothesis isn't literally true in all cases. But it is true in nearly all cases, as Torvalds has been in an excellent position to see.

Re:Whose Eyes?

[Your.Master]

Torvald's didn't say the "many eyes" thing at all. Eric S. Raymond did.

Re:The best bug is the one not written

[Gaygirlie]

Software can't be made secure afterwards, it must be the the primary goal.

That's bullshit. Software can definitely be made secure afterwards even if it wasn't that to begin with, there is no other obstacle to that than manpower and time. Also, security being a primary goal does not guarantee that there won't be bugs, so again, that makes that saying utterly ignorant. Bugs, by very definition, are accidental issues, not designed-in features, and no amount of "primary goals" will guarantee that mistakes and accidents won't happen.

Second Linus Law: Curse the bugs out

[JoeyRox]

Maybe Linus isn't cursing at the developers with enough frequency or intensity?

Re:Second Linus Law: Curse the bugs out

[Kjella]

Maybe Linus isn't cursing at the developers with enough frequency or intensity?

It seems the kernel is rarely the problem, so I'd say the amount of cursing is just right. The problem is Linus doesn't run all these other projects.

Re-engineer the OS to include ROMs?

[Redfearn]

Is there a way to re-engineer operating systems so that some parts are strictly read-only (like, baked in ROM chips); other parts difficult to change (flash them?), and so on? Right now, it seems all data, programs and operating system components are equally vulnerable to writes by viruses. How many people would be harmed if some basic components of XP had been burned into ROM? Then anti-virus programs could hook into those "fortified" modules to maitain or restore the integrity of other parts.

Re:Modern software...

[Kjella]

Doesn't that just about say it all? More eyes don't solve complexity issues, only more brains and better architecture.

I think that if you do some research - at least if you limit yourself to human subjects - you will find there's a strong correlation between number of eyes and number of brains so "more eyes" implies "more brains". And if you can settle the age-old discussing of whether encapsulation, abstractions and design patterns reduce or increase complexity you should the IT Peace Prize.

Re:Too late...

[Kjella]

By the time something becomes "core infrastructure", it's usually not in a condition where a rewrite is at all advisable. You have an existing code base that's seen lots of real world usage and presumably works well most of the time, what you need is testing, cleanup, sanity-checking, error handling and formal verification that it performs as intended. And it's particularly important that you review obscure functionality like the heartbeat TLS extension that lead to the heartbleed bug, that you put many eyes were few have wandered before.

Of course if you find that something's genuinely missing, like a layer to prevent SQL injection and just depending on every piece of code doing the "right thing" that might be a reason to re-architect, but I think your advice is far more applicable to new development than the projects were talking about here. For example there's nothing wrong with the heartbeat spec, it's fine. It's the implementation that was fatally flawed and the only way you can catch that is reviewing the code.