Samuel Thibault

Most people don't realize it, but the Hurd system is actually well established. About 75% of Debian official packages do build fine, it has mainstream gcc/glibc/llvm support, go and rust ports are ongoing, it can be installed with the Debian installer and GuixSD and Arch ports are ongoing...

Yet not so much has been happening within the Hurd itself in the past couple of years. We have notably added a PCI arbiter, which allows for both flexible and safe PCI access for end users, and some basic ACPI support is ongoing. But many exciting features could be achieved with a bit of work.

This talk will discuss some of these promising features, to give a sort of ideas roadmap for contributions. Some have implementation sketches which just need to be polished to be more production-ready, such as httpfs, mboxfs, or writing translators in more high-level languages than C. Other features are at early stage, such as adding sound support through rump, getting complete rid of disk drivers from the kernel by moving them to userland, or also getting valgrind support. I will also discuss some promising ideas, such as using rump to get support for more filesystems.

Accessibility in free software has been much improved between 2004 and 2009 thanks to Sun, which massively invested for it. Since it has disappeared, big free software projects have experienced many regressions in this matter. Some of their maintainers consider that there is not enough money to deal with the matter.

Given the complexity of code such as Libreoffice, and given the status of accessibility of software in modern society, it becomes impossible to invest millions of euros to pay for someone who would fix current bugs and track future ones in millions of code lines, through hundreds of commits, and permanently. All the more since beyond this task, a paid expert should rather work about improvements of assistive technologies themselves and pipes (at-spi, atk), instead of fixing miscoding of unaware developers. So, to hope a free and accessible software, major projects need to take responsibility of this. Non-regression tests could be a way to make this duty lighter, even if not a complete solution. What kind of tests could be set?

One of Hurd's goals is to isolate hardware drivers, to provide flexibility and safety, by isolating drivers' possibly erratic behavior, delegating driving hardware to unprivileged users, etc.

We are thus currently working on a PCI arbiter, which allows to let different userland drivers access PCI devices concurrently, while paving the way for fine-grain per-user, per-session, etc. permission management over PCI access, the use of IOMMUs allowing to do that very safely.

The Debian accessibility team aims at making Debian work nicely for people with disabilities. This of course means packaging accessibility-related software and bringing accessibility fixes in, but it also means integrating notions of accessibility into all stages of Debian, from d-i bootup to bug reporting. We will discuss past achievements and challenges ahead, which everybody can help with.

The Hurd recently saw some experimental work to add sound support. This was done thanks to the Rump kernel, embedded in a library. As a first-working implementation the library is directly linked into mplayer, but of course we plan to make this much more flexible by moving the Rump sound driver into a separate translator, very much like we did for network drivers with netdde. This however poses design questions. The other hardware bit missing on the Hurd is USB support, which we should be able to add a similar way. This poses yet more design questions, as to the shape of the hird of translators we should start for this. This will of course also be the opportunity to give the now-traditional yearly latest news about the Hurd.

All users should be able to use all applications, despite any kind of condition (blindness, colorblindness, deafness, ...). Without accessibility, software excludes the people who can not use them. I will introduce accessibility, the accessibility stack, and the good practices to keep in mind while designing applications so they get accessible by design.

All users should be able to use all applications, despite any kind of condition (blindness, colorblindness, deafness, ...). Without accessibility, software excludes the people who can not use them. Fixing the accessibility of an application after having developped it can be quite costly, while designing applications with accessibility in mind from the start does not cost much.

Actually, for a fair part, accessibility in design boils down to well-known good practice, such as structuring the application logically without any particular rendering in mind.

I will introduce to accessibility, the accessibility stack, and the good practices to keep in mind while designing applications so they get accessible by design.

It is not so obvious to newcomers to the Hurd project how to follow execution path from their application to actual behavior in the Hurd translators. This talk will guide them through the involved RPCs, thus providing a starting guide into the GNU Hurd architecture, and hopefully, to contributing fixes or new features!

With monolithic kernels, the path from application source code to actual implementation in the kernel is very simple: only the system call layer needs to be understood. In a multiserver system such as the GNU Hurd, this is not so simple, since one for a start has to understand which server one is actually talking to, and how RPCs work. Once this is cleared up, it becomes way easier to know where to contribute fixes and features!

This talk will thus take a few examples of typical application operations (opening and reading a file, opening and working with a socket) and go through the RPC path, to illustrate how translators are involved and where the eventual implementations actually reside.

It will also be an opportunity to give the latest news about the GNU/Hurd project.

This will discuss the few important details that distribution hackers should know to make sure that their distribution is accessible to everybody.

Freedom #0 of free software is to be able to execute the software. But that is useless if one is unable to actually use the software. Distributions should thus take particular care of their accessibility, which means shipping accessibility tools of course, but also making sure that some accessible software alternatives are packaged, that accessibility can easily be enabled on an installed system, and even that the distribution installer itself is accessible!

Thanks to the experience we have had in Debian in the past years, I will guide the audience through the few important details to know for distribution hackers in order to make sure that their distribution is accessible.

We will explain how userland drivers are implement in GNU/Hurd thanks to the DDE layer, and what kernel support is needed for that. We will also show the flexibility this brings for the user.

A longer version of this talk (incl video) is also available at https://www.gnu.org/ghm/2013/paris/

The podium will consist of representatives of all the OS projects participating in the devroom. This is thought as an interactive event. The audience will have the opportunity to ask questions, which will be answered by all representatives.

Other people: Martin Děcký Norman Feske Stefan Kalkowski Björn Döbel Nils Asmussen

GNU/Hurd aims at being a general-purpose Operating System with a strong emphasis on flexibility and freedom for the user, and thus based on a design made of a micro-kernel surrounded by a hurd of userland servers. It has however a long-term vaporware reputation. Development has indeed been relatively dormant for some time, but recent regain of interest has brought interesting improvements and stabilization, to the point that there will be a non-official release of the GNU/Hurd variant of Debian Wheezy, with about 75% of the Debian packages, including classical graphical desktop application (gnome, gnumeric, firefox, ...)

This talk will present to GNU/Hurd in general and its "translator" mechanism which replaces the traditional notion of filesystem by userland processes, so as to provide strong flexibility to users and administrators, and we will demo it live. The subhurd/neighborhurd mechanism, a very natural way to provide virtualization container support on GNU/Hurd, will also be presented. We will also present recent developments, notably in terms of DDE device drivers run as userland processes, and discuss about maintenance of DDE.