Last week the Flatpak community woke to the “news” that we are making the world a less secure place and we need to rethink what we’re doing. Personally, I’m not sure this is a fair assessment of the situation. The “tl;dr” summary is: Flatpak confers many benefits besides the sandboxing, and even looking just at the sandboxing, improving app security is a huge problem space and so is a work in progress across multiple upstream projects. Much of what has been achieved so far already delivers incremental improvements in security, and we’re making solid progress on the wider app distribution and portability problem space.

Sandboxing, like security in general, isn’t a binary thing – you can’t just say because you have a sandbox, you have 100% security. Like having two locks on your front door, two front doors, or locks on your windows too, sensible security is about defense in depth. Each barrier that you implement precludes some invalid or possibly malicious behaviour. You hope that in total, all of these barriers would prevent anything bad, but you can never really guarantee this – it’s about multiplying together probabilities to get a smaller number. A computer which is switched off, in a locked faraday cage, with no connectivity, is perfectly secure – but it’s also perfectly useless because you cannot actually use it. Sandboxing is very much the same – whilst you could easily take systemd-nspawn, Docker or any other container technology of choice and 100% lock down a desktop app, you wouldn’t be able to interact with it at all.

Network services have incubated and driven most of the container usage on Linux up until now but they are fundamentally different to desktop applications. For services you can write a simple list of permissions like, “listen on this network port” and “save files over here” whereas desktop applications have a much larger number of touchpoints to the outside world which the user expects and requires for normal functionality. Just thinking off the top of my head you need to consider access to the filesystem, display server, input devices, notifications, IPC, accessibility, fonts, themes, configuration, audio playback and capture, video playback, screen sharing, GPU hardware, printing, app launching, removable media, and joysticks. Without making holes in the sandbox to allow access to these in to your app, it either wouldn’t work at all, or it wouldn’t work in the way that people have come to expect.

What Flatpak brings to this is understanding of the specific desktop app problem space – most of what I listed above is to a greater or lesser extent understood by Flatpak, or support is planned. The Flatpak sandbox is very configurable, allowing the application author to specify which of these resources they need access to. The Flatpak CLI asks the user about these during installation, and we provide the flatpak override command to allow the user to add or remove these sandbox escapes. Flatpak has introduced portals into the Linux desktop ecosystem, which we’re really pleased to be sharing with snap since earlier this year, to provide runtime access to resources outside the sandbox based on policy and user consent. For instance, document access, app launching, input methods and recursive sandboxing (“sandbox me harder”) have portals.

The starting security position on the desktop was quite terrible – anything in your session had basically complete access to everything belonging to your user, and many places to hide.

• Access to the X socket allows arbitrary input and output to any other app on your desktop, but without it, no app on an X desktop would work. Wayland fixes this, so Flatpak has a fallback setting to allow Wayland to be used if present, and the X socket to be shared if not.
• Unrestricted access to the PulseAudio socket allows you to reconfigure audio routing, capture microphone input, etc. To ensure user consent we need a portal to control this, where by default you can play audio back but device access needs consent and work is under way to create this portal.
• Access to the webcam device node means an app can capture video whenever it wants – solving this required a whole new project.
• Sandboxing access to configuration in dconf is a priority for the project right now, after the 1.0 release.

Even with these caveats, Flatpak brings a bunch of default sandboxing – IPC filtering, a new filesystem, process and UID namespace, seccomp filtering, an immutable /usr and /app – and each of these is already a barrier to certain attacks.

Looking at the specific concerns raised:

• Hopefully from the above it’s clear that sandboxing desktop apps isn’t just a switch we can flick overnight, but what we already have is far better than having nothing at all. It’s not the intention of Flatpak to somehow mislead people that sandboxed means somehow impervious to all known security issues and can access nothing whatsoever, but we do want to encourage the use of the new technology so that we can work together on driving adoption and making improvements together. The idea is that over time, as the portals are filled out to cover the majority of the interfaces described, and supported in the major widget sets / frameworks, the criteria for earning a nice “sandboxed” badge or submitting your app to Flathub will become stricter. Many of the apps that access --filesystem=home are because they use old widget sets like Gtk2+ and frameworks like Electron that don’t support portals (yet!). Contributions to improve portal integration into other frameworks and desktops are very welcome and as mentioned above will also improve integration and security in other systems that use portals, such as snap.
• As Alex has already blogged, the freedesktop.org 1.6 runtime was something we threw together because we needed something distro agnostic to actually be able to bootstrap the entire concept of Flatpak and runtimes. A confusing mishmash of Yocto with flatpak-builder, it’s thankfully nearing some form of retirement after a recent round of security fixes. The replacement freedesktop-sdk project has just released its first stable 18.08 release, and rather than “one or two people in their spare time because something like this needs to exist”, is backed by a team from Codethink and with support from the Flatpak, GNOME and KDE communities.
• I’m not sure how fixing and disclosing a security problem in a relatively immature pre-1.0 program (in June 2017, Flathub had less than 50 apps) is considered an ongoing problem from a security perspective. The wording in the release notes?

Zooming out a little bit, I think it’s worth also highlighting some of the other reasons why Flatpak exists at all – these are far bigger problems with the Linux desktop ecosystem than app security alone, and Flatpak brings a huge array of benefits to the table:

• Allowing apps to become agnostic of their underlying distribution. The reason that runtimes exist at all is so that apps can specify the ABI and dependencies that they need, and you can run it on whatever distro you want. Flatpak has had this from day one, and it’s been hugely reliable because the sandboxed /usr means the app can rely on getting whatever they need. This is the foundation on which everything else is built.
• Separating the release/update cadence of distributions from the apps. The flip side of this, which I think is huge for more conservative platforms like Debian or enterprise distributions which don’t want to break their ABIs, hardware support or other guarantees, is that you can still get new apps into users hands. Wider than this, I think it allows us huge new freedoms to move in a direction of reinventing the distro – once you start to pull the gnarly complexity of apps and their dependencies into sandboxes, your constraints are hugely reduced and you can slim down or radically rethink the host system underneath. At Endless OS, Flatpak literally changed the structure of our engineering team, and for the first time allowed us to develop and deliver our OS, SDK and apps in independent teams each with their own cadence.
• Disintermediating app developers from their users. Flathub now offers over 400 apps, and (at a rough count by Nick Richards over the summer) over half of them are directly maintained by or maintained in conjunction with the upstream developers. This is fantastic – we get the releases when they come out, the developers can choose the dependencies and configuration they need – and they get to deliver this same experience to everyone.
• Decentralised. Anyone can set up a Flatpak repo! We started our own at Flathub because there needs to be a center of gravity and a complete story to build out a user and developer base, but the idea is that anyone can use the same tools that we do, and publish whatever/wherever they want. GNOME uses GitLab CI to publish nightly Flatpak builds, KDE is setting up the same in their infrastructure, and Fedora is working on completely different infrastructure to build and deliver their packaged applications as Flatpaks.
• Easy to build. I’ve worked on Debian packages, RPMs, Yocto, etc and I can honestly say that flatpak-builder has done a very good job of making it really easy to put your app manifest together. Because the builds are sandboxed and each runtimes brings with it a consistent SDK environment, they are very reliably reproducible. It’s worth just calling this out because when you’re trying to attract developers to your platform or contributors to your app, hurdles like complex or fragile tools and build processes to learn and debug all add resistance and drag, and discourage contributions. GNOME Builder can take any flatpak’d app and build it for you automatically, ready to hack within minutes.
• Different ways to distribute apps. Using OSTree under the hood, Flatpak supports single-file app .bundles, pulling from OSTree repos and OCI registries, and at Endless we’ve been working on peer-to-peer distribution like USB sticks and LAN sharing.

Nobody is trying to claim that Flatpak solves all of the problems at once, or that what we have is anywhere near perfect or completely secure, but I think what we have is pretty damn cool (I just wish we’d had it 10 years ago!). Even just in the security space, the overall effort we need is huge, but this is a journey that we are happy to be embarking together with the whole Linux desktop community. Thanks for reading, trying it out, and lending us a hand.

At the Gtk+ hackfest in London earlier this year, we stole an afternoon from the toolkit folks (sorry!) to talk about Flatpak, and how we could establish a “critical mass” behind the Flatpak format. Bringing Linux container and sandboxing technology together with ostree, we’ve got a technology that solves real world distribution, technical and security problems that have arguably held back the Linux desktop space and frustrated ISVs and app developers for nearly 20 years. The problem we need to solve, like any ecosystem, is one of users and developers – without stuff you can easily get in Flatpak format, there won’t be many users, and without many users, we won’t have a strong or compelling incentive for developers to take their precious time to understand a new format and a new technology.

As Alex Larsson said in his GUADEC talk yesterday: Decentralisation is good. Flatpak is a tool that is totally agnostic of who is publishing the software and where it comes from. For software freedom, that’s an important thing because we want technology to empower users, rather than tell them what they can or can’t do. Unfortunately, decentralisation makes for a terrible user experience. At present, the Flatpak webpage has a manually curated list of links to 10s of places where you can find different Flatpaks and add them to your system. You can’t easily search and browse to find apps to try out – so it’s clear that if the current situation remains we’re not going to be able to get a critical mass of users and developers around Flatpak.

Enter Flathub. The idea is that by creating an obvious “center of gravity” for the Flatpak community to contribute and build their apps, users will have one place to go and find the best that the Linux app ecosystem has to offer. We can take care of the boring stuff like running a build service and empower Linux application developers to choose how and when their app gets out to their users. After the London hackfest we sketched out a minimum viable system – Github, Buildbot and a few workers – and got it going over the past few months, culminating in a mini-fundraiser to pay for the hosting of a production-ready setup. Thanks to the 20 individuals who supported our fundraiser, to Mythic Beasts who provided a server along with management, monitoring and heaps of bandwidth, and to Codethink and Scaleway who provide our ARM and Intel workers respectively.

We inherit our core principles from the Flatpak project – we want the Flatpak technology to succeed at alleviating the issues faced by app developers in targeting a diverse set of Linux platforms. None of this stops you from building and hosting your own Flatpak repos and we look forward to this being a wide and open playing field. We care about the success of the Linux desktop as a platform, so we are open to proprietary applications through Flatpak’s “extra data” feature where the client machine downloads 3rd party binaries. They are correctly labeled as such in the AppStream, so will only be shown if you or your OS has configured GNOME Software to show you apps with proprietary licenses, respecting the user’s preference.

The new infrastructure is up and running and I put it into production on Thursday. We rebuilt the whole repository on the new system over the course of the week, signing everything with our new 4096-bit key stored on a Yubikey smartcard USB key. We have 66 apps at the moment, although Alex is working on bringing in the GNOME apps at present – we hope those will be joined soon by the KDE apps, and Endless is planning to move over as many of our 3rd party Flatpaks as possible over the coming months.

So, thanks again to Alex and the whole Flatpak community, and the individuals and the companies who supported making this a reality. You can add the repository and get downloading right away. Welcome to Flathub! Go forth and flatten… 🙂

Like all good blog posts, this one starts with an apology about not blogging for ages – in my case it looks like it’s been about 7 years which is definitely a new personal best (perhaps the equally or more remarkable thing is that I have diligently kept WordPress running in the meantime). In that time, as you might expect, a few things have happened, like I met a wonderful woman and fell in love and we have two wonderful children. I also decided to part ways with my “first baby” and leave my role as CTO & co-founder of Collabora. This was obviously a very tough decision – it’s a fantastic team where I met and made many life-long friends, and they are still going strong and doing awesome things with Open Source. However, shortly after that, in February last year, I was lucky enough to land what is basically a dream job working at Endless Computers as the VP of Deployment.

As I’m sure most readers know, Endless is building an OS to bring personal computing to millions of new users across the world. It’s based on Free Software like GNOME, Debian, ostree and Flatpak, and the more successful Endless is, the more people who get access to education, technology and opportunity – and the more FOSS users and developers there will be in the world. But in my role, I get to help define the product, understand our users and how technology might help them, take Open Source out to new people, solve commercial problems to get Endless OS out into the world, manage a fantastic team and work with bright people, learn from great managers and mentors, and still find time to squash bugs, review and write more code than I used to. Like any startup, we have a lot to do and not enough time to do it, so although there aren’t quite enough days in the week, I’m really happy!

In any case, the main point of this blog post is that I’m at GUADEC in Manchester right now, and I’d like to blog about Flathub, but I thought it would be weird to just show up and say that after 7 years of silence without saying hello again. 🙂

As you all know by now, exciting moves from Google on the WebM project have lead to them open-sourcing On2’s VP8 codec to provide a freely available video codec for HTML5 content. Collabora Multimedia worked with Entropy Wave to add support to GStreamer for the new codec from day 1, and I was really happy yesterday to update my Debian system and get the support installed locally too. Thanks to our and Igalia’s fine work on GStreamer HTML5 support in WebKitGTK+, Gustavo Noronha found it worked out of the box with Epiphany too.

Predictably, the MPEG-LA aren’t too pleased with this, and are no doubt winding up their PR and industry allies at the moment, as well as this opening a new front on the Apple vs Google ongoing platform battle. But if your business model is collecting money through what is essentially a protection racket and spreading FUD about patent litigation, the VP8 license implicitly creating a zero-cost zero-revenue patent pool is not going to be good news for you (from the department of Google deleting your business model). The question is now whether the allure of Google’s content will win over against the legal chest pounding of the patent trolls, and whether they start flipping switches to make YouTube only serve up WebM content after a while.

Also in amazing and incredible news, Collabora’s Telepathy/GStreamer/GNOME/Debian/general R&D guru and staunch Web 2.0 holdout Sjoerd Simons has actually now got a blog after a mere 3 years of us suggesting it to him since he joined Collabora as an intern. He’s been hacking on some RTP payloader elements for VP8 so we can use it for video calling on the free desktop. All very exciting stuff, especially in conjunction with Muji (multi-user video calls over XMPP) support heading into Telepathy thanks to NLNet‘s ongoing support.

The first Telepathy session session on Saturday evening at the Boston GNOME Summit was very much of a Q&A where myself and Will answered various technical and roadmap issues from a handful of developers and downstream distributors. It showed me that there’s a fair amount of roadmap information we should do better at communicating outside of the Telepathy project, so in the hope its useful to others, read on…

(more…)

I spent this weekend in Boston for the annual GNOME summit. I really enjoyed it this year, although there were fewer attendees than previously it felt very focussed and productive. There’s some cool stuff going on, and it’s always great to catch up with all of the usual free software suspects in Boston. Some highlights from the weekend:

• Corridor session with David Zeuthen, Ryan Lortie, Matthias Clasen and later joined by me and Will Thompson from Collabora, discussed a lot of the issues of integrating D-Bus into Glib and I think achieved a pretty good consensus about how GVariant and GDBus should fit together and start getting the pieces merged. Really looking forwards to it, GVariant looks mad ninja.
• Some nice discussions about Gtk+ 3.0 roadmap, although I’m still worried that the sealing/accessor work will take so much developer time there won’t be that much time to make the improvements its supposed to enable. Theming was mentioned but what else should there be? Also spotted Kristian Høgsberg, Cody Russell and Matthias again talking about client-side decorations, presumably Wayland scheming… 🙂
• Nice session about messaging and other notifications in GNOME Shell. I really like the way the project is going, and hopefully we can join in and spend some time hooking Telepathy up here, although as a fallback to make the existing stuff (which doesn’t have specific code to hook in and make a UI), then it’d be nice if the shell could also be a frontend for the notify/indicate stuff which Ubuntu have been pushing in GNOME 2.x already.
• Had some good brainstorming about Telepathy integration in games and Tomboy sharing too, as well as some slightly less conclusive pondering about how to deliver metacontacts (ie, merging multiple sources such as IM, social networking and other address books) in GNOME. We need to sync up with what Moblin is doing here as it looks quite promising and should give us some components to re-use.
• Will, Sjoerd and I spent an hour or two around a blackboard working out how XMPP end-to-end TLS encryption might be exposed to clients in Telepathy, in order to work out how best we’d expose OTR too. It looks like we have a fairly workable proposal now which we’ll be explaining in due course, but it means at least we can give more useful advice to people who are interested in implementing it, or move forwards on implementing it ourselves.

I was really impressed by Jason Clinton and others’ summaries of the sessions, which I think are really valuable for the people who couldn’t make it to the summit. He asked me to take some notes about the first Telepathy session on Saturday evening while he was taking notes about the Outreach session. Rather than lumber him with my deranged scratchings from Tomboy, I’ll blog them separately.

If Collabora has seemed quiet recently, it’s because we’ve been slaving away on various parts of a really awesome project, which we can finally start talking about. Maemo 5 is coming! o o/ /o/

Flyby blog entry. Been an awesome week, hope people enjoyed the Collabora party last night. Getting to the GNOME Shell BOF this morning was a real struggle, but I really like the way its going. We’ve scheduled an Empathy BOF for Friday at 15:45 in room 2-4, hoping to talk with folks about improving the accounts UI, tracker/addressbook and GNOME Shell integration.

Aside from that, we’re planning to hack on Empathy for basically the whole day, looking at MC5 porting and UI polish, so there will be a big load of Collabora folks and friends somewhere. Will update when I know where.

I’ve been very impressed several times in the past few months when I’ve discovered awesome new top-like utilities. I’m probably being slow on the uptake and everyone else but me knows about these, but in case its not just me thats been stuck in the ’70s:

htop

A much-needed refresh of oldschool top, this still works on your beloved console but gives you visual bar-graphs of CPU, RAM and swap, lets you scroll through the processes and deliver signals/renicing without having to copy the PID off the moving target. Its like the future!

iftop

One of those things I use so often now I have no idea how I even survived without it. Why is this server lagging, who’s hogging the wireless/DSL, which VM is chewing all of the upstream bandwidth? iftop shows you at a glance how much traffic is being used by which host pairs on a given interface, and you can toggle port numbers on and off with simple key-presses. Absolutely indispensable.

iotop

Does this box feel slow to anyone else? Is it swapping, or is it the database server chewing all the IO? Why does my drive keep seeking? It’s amazing… top for IO bandwidth usage!

A passing mention is deserved for apachetop too, which is pretty neat, but when a server is being hammered it’s not something I found too hard to get a feel for just by tailing the log for a while, so it’s not been as life-changing as the others. Maybe that just means my servers don’t see enough traffic.

“… all programs that interact with e-mail are broken in one way or another. Please be careful.” – Lars Wirzenius

I seem to have a cunning knack of finding problems with configuring server software, particularly involving e-mail, where a) I can’t find answers in Google, b) most people I go and ask for help say they’d usually ask me such things, and c) if I go onto IRC or mailing lists I end up helping other people and not getting any help with my problem. It’s quite likely this is just because I’m something of a perfectionist, so the ridiculous crappy hacks people come up with and seem content to entrust their mail to are unacceptable to me for one reason or another. Anyway, in my ongoing quest for the perfect mail system, I’ve painted myself into a corner again.

(I’m currently running with postfix, postgrey, clamav-milter, dspam, dovecot using LDA, managesieve and the cmusieve & antispam plugins. If I can get the current incarnation working, I’ve had enough requests to write up a full HOWTO, and seen enough around with pretty questionable content, that I’ll probably do it before too long.)

I’ve got postfix’s local transport configured to hand mail to dspam over LMTP, using mailbox_transport = lmtp:unix:/dspam/dspam.sock. dspam is configured to listen here, add X-DSPAM-Result and signature headers, and then deliver the mail with dovecot’s deliver LDA (which I’ve set to 4750 root:dspam). From dspam.conf:

ServerParameters "--deliver=innocent,spam -d %u"
ServerDomainSocketPath "/var/spool/postfix/dspam/dspam.sock"
...
Preference "spamAction=tag"
Preference "signatureLocation=headers"
...
TrustedDeliveryAgent "/usr/lib/dovecot/deliver"

My dovecot configuration is pretty standard, using PAM for both passdb and userdb, and provides the auth-master socket that deliver needs. The problem I have is that postfix’s local transport is qualifying the local username with the FQDN of the machine before delivering it to dspam with LMTP (the local mail transfer protocol), even for locally-originated mail which was only addressed with a bare username! dspam doesn’t mangle it or care if the user is local or not, and then cheerfully invokes deliver -d robot101@omega.example.co.uk, which returns EX_NOUSER (addressee unknown) because my username is just robot101. From mail robot101:

Mar 6 02:25:40 omega postfix/pickup[13607]: DAA4942F41F: uid=1000 from=<robot101>
Mar 6 02:25:40 omega postfix/cleanup[13637]: DAA4942F41F: message-id=<20090306022540.DAA4942F41F@omega.example.co.uk>
Mar 6 02:25:40 omega postfix/qmgr[12552]: DAA4942F41F: from=<robot101@omega.example.co.uk>, size=339, nrcpt=1 (queue active)
Mar 6 02:25:40 omega dovecot: auth(default): passwd(robot101@omega.example.co.uk): unknown user
Mar 6 02:25:40 omega dspam[13527]: Delivery agent returned exit code 67: /usr/lib/dovecot/deliver -d robot101@omega.example.co.uk
Mar 6 02:25:40 omega postfix/lmtp[13640]: DAA4942F41F: to=<robot101@omega.example.co.uk>, orig_to=<robot101>, relay=omega.example.co.uk[/dspam/dspam.sock], delay=0.08, delays=0.05/0.01/0.01/0.03, dsn=4.3.0, status=deferred (host omega.example.co.uk[/dspam/dspam.sock] said: 421 4.3.0 <robot101@omega.example.co.uk> Delivered (in reply to end of DATA command))

So, no e-mail for me. Dearest lazyweb, which of the three components is behaving wrongly, and how can I fix it?

(And no, I’m not just going to switch to GMail. I store my data on hard drives, which are sometimes in servers, not “in the cloud”. Until about a month ago, most people I knew spoke about clouds which were made of particles of water in the sky, rather than as a data storage media. What if it rains? 😉

Update: The problem is fixed! Even though arguably the problem is dspam’s for not knowing which users are local or not, it’s fixable in dovecot 1.1.x using the auth_username_format = %n option. Thanks so much to Angel Marin for helping me out.

Update 2: There’s also a patch for dspam floating around which adds a StripRcptDomain option, which makes the LMTP server truncate the e-mail address at the @, so essentially assumes everyone to be a local user. The problem with both of these fixes is that they’re both blunt instruments which will break virtual users on the same host. I think the real fix would be something more like a LocalDomains option in dspam, to choose which domains should be considered local and truncated from the e-mail addresses for delivery purposes.