Commit 7b4deec4 authored by anonym's avatar anonym
Browse files

Merge remote-tracking branch 'origin/master' into stable

parents 3a886265 36d3479e
......@@ -225,7 +225,7 @@ po_slave_languages:
- fr|Français
- pt|Português
# PageSpec controlling which pages are translatable
po_translatable_pages: '!security/audits and !security/audits/* and (about or bugs or chat or contribute or contribute/how/donate or doc or doc/* or download or getting_started or inc/release_notes/* or inc/stable_i386_date or inc/stable_i386_release_notes or inc/stable_i386_iso_size_*_unit or index or news or news/* or press or press/* or security or security/* or sidebar or support or support/* or todo or torrents or wishlist or misc or misc/*)'
po_translatable_pages: '!security/audits and !security/audits/* and (about or bugs or chat or contribute or contribute/how/donate or doc or doc/* or download or getting_started or inc/release_notes/* or inc/stable_i386_date or inc/stable_i386_release_notes or inc/stable_i386_iso_size_mb_unit or inc/stable_i386_iso_size_mib_unit or index or news or news/* or press or press/* or security or security/* or sidebar or support or support/* or todo or torrents or wishlist or misc or misc/*)'
# internal linking behavior (default/current/negotiated)
po_link_to: current
......
......@@ -202,7 +202,7 @@ po_slave_languages:
- fr|Français
- pt|Português
# PageSpec controlling which pages are translatable
po_translatable_pages: '!security/audits and !security/audits/* and (about or bugs or chat or contribute or contribute/how/donate or doc or doc/* or download or getting_started or inc/release_notes/* or inc/stable_i386_date or inc/stable_i386_release_notes or inc/stable_i386_iso_size_*_unit or index or news or news/* or press or press/* or security or security/* or sidebar or support or support/* or todo or torrents or wishlist or misc or misc/*)'
po_translatable_pages: '!security/audits and !security/audits/* and (about or bugs or chat or contribute or contribute/how/donate or doc or doc/* or download or getting_started or inc/release_notes/* or inc/stable_i386_date or inc/stable_i386_release_notes or inc/stable_i386_iso_size_mb_unit or inc/stable_i386_iso_size_mib_unit or index or news or news/* or press or press/* or security or security/* or sidebar or support or support/* or todo or torrents or wishlist or misc or misc/*)'
# internal linking behavior (default/current/negotiated)
po_link_to: current
......
......@@ -38,8 +38,28 @@ SHOULD
- Replace the list of bad users. Flag them as nasty automatically
- Allow forwarding issues from and to other user support projects (Tor, Access Now)
- Allow users to express whether they were satisfied with our answers
- Be configurable using Puppet
MAY
---
- Have a disposable chat system for tricky cases (Tor does that)
Resources
=========
- [[!wikipedia Comparison_of_help_desk_issue_tracking_software]] (Wikipedia)
Options
=======
### OTRS
- http://www.otrs.com/
- https://otrs.github.io/doc/manual/admin/3.1/en/html/configure-pgp.html
### RT
- http://bestpractical.com/rt/
- https://www.bestpractical.com/docs/rt/4.2/RT/Crypt/GnuPG.html
- https://forge.puppetlabs.com/darin/rt
* Needs GLib 2.40 (in Jessie).
* Intro: <https://wiki.gnome.org/HowDoI/GNotification>
* "applications using GNotification should be able to be started as
a D-Bus service, using GApplication." =>
<https://wiki.gnome.org/HowDoI/DBusApplicationLaunching>
* "gnome-shell uses desktop files to find extra information (app icon,
name) about the sender of the notification. If you don't have
a desktop file whose base name matches the application id, then your
notification will not show up". We'll see.
* If it's deemed overkill, or just too painful, to port every single
script that needs to send notifications with actions to
"GApplication + D-Bus activation + actions support" properly,
a solution could be to let them all pretend to be the
`org.boum.tails.OpenURL` or similar app in the simplest possible
way, i.e. create a GApplication object merely to be able to send the
notification or use the private D-Bus API to
`org.gtk.Notifications`.
And then, we need only one single real GApplication that can react
to actions with parameters.
A potential problem with this approach is that if a given
GApplication is running already, another process pretending to be
the same can't send notifications (reproducer code follows; the
private D-Bus API allows to workaround this problem, but has other
ones, read on). So, in practice if we go this way the URL opener app
must have a lifetime close to zero, otherwise when clicking on
multiple notifications buttons in a row, some actions won't be
triggered (i.e. URLs won't be opened).
#!/usr/bin/perl
use strict;
use warnings FATAL => 'all';
use 5.10.1;
use Gtk3;
use Glib::Object::Introspection;
use Try::Tiny;
Glib::Object::Introspection->setup(
basename => 'Gio',
version => '2.0',
package => 'Gio'
);
my $app_id = 'org.gnome.gedit';
my $app = Gtk3::Application->new($app_id, []);
$app->register() || die "Could not register app";
my $notification = Gio::Notification->new('My notification title');
$notification->set_body("My notification body");
$notification->add_button("OK", 'app.new-document');
# If gedit is already running, this fails with:
# GLib-GIO-CRITICAL **: g_application_send_notification:
# assertion '!g_application_get_is_remote (application)' failed
$app->send_notification("my-notification-id", $notification)
or die "Could not send notification";
* private D-Bus API (`org.gtk.Notifications`):
<https://wiki.gnome.org/Projects/GLib/GNotification> ... maybe an
option whenever language bindings are not good enough, but so far
I (intrigeri) was not able to add buttons to it:
#!/usr/bin/perl
use strict;
use warnings FATAL => 'all';
use 5.10.1;
use Net::DBus;
Net::DBus
->session
->get_service("org.gtk.Notifications")
->get_object("/org/gtk/Notifications")
->as_interface("org.gtk.Notifications")
->AddNotification('org.gnome.gedit', 'whatever-notification-id', {
title => 'my notification title',
body => 'my notification body',
# Works... when 'buttons' is not passed
'default-action' => 'app.new-window',
# Does not work, for some reason
buttons => [
{ label => "New window", action => "app.new-window"},
{ label => "New document", action => "app.new-document"},
],
});
......@@ -62,8 +62,55 @@ Tools to manage containers
Running GUI applications in containers
======================================
We're told that Subgraph OS will use LXC + xpra. In this vein:
<a id="oz"></a>
## Subgraph's Oz
* [Oz](https://github.com/subgraph/oz) is a sandboxing system targeting
everyday workstation applications. It relies mainly on Xpra and
Linux namespaces.
* See its [technical
documentation](https://github.com/subgraph/oz/wiki/Oz-Technical-Details)
for details.
* It has a [GNOME Shell
extension](https://github.com/subgraph/ozshell-gnome-extension) that
e.g. allows one to add files (read-only or read-write) to a running sandbox.
* Sandboxed applications can optionally be given access to:
- files passed on their command line (in a read-only manner), which
makes them work just fine when started e.g. from a file manager or
a web browser;
- the microphone and/or speaker, thanks to Xpra.
* Potential functionality limitations and UX issues:
- It's not clear to me how things such as copy'n'pasting between
applications, printing, Orca and IBus input methods are handled.
Some of those need access to the session D-Bus, so
presumably making these work will require mediating D-Bus calls,
as done e.g. in AppArmor on Ubuntu.
* Xpra 0.14+ can be set up to configure IBus, but how does this
integrate with the IBus configuration changes applied via the
desktop environment?
* Xpra 0.15+ provides a CUPS forwarder backend
- When one has already started e.g. LibreOffice and needs to open
a new file, they need to first use the GNOME Shell extension to
add it to the sandbox; in this respect, the resulting UX seems to
be inferior to the one the designs the AppArmor and `xdg-app`
folks mean to implement (file chooser with a privileged backend).
Creating new files might be problematic as well.
- Most host OS directories are bind-mounted (read-only) inside the
sandbox, and a blacklist is applied on top. This means that
arbitrary code execution is possible, but confined within the
sandbox, while with AppArmor we restrict a bit more what external
programs can be executed, but the equivalent of the sandbox is
arguably weaker.
- Xpra doesn't support Wayland yet, and Wayland will make some of
Xpra's current advantages obsolete.
## Other GUI in containers leads
* [Subuser](http://subuser.org/): essentially Docker + Xpra
* [GNOME sandboxed
applications](https://wiki.gnome.org/Projects/SandboxedApps), aka.
`xdg-app`; their concept of "portals" is very interesting.
* <http://pleonasm.info/blog/2012/10/privilege-separation-with-xpra/>
* [docker-desktop](https://github.com/rogaha/docker-desktop)
* Stéphane Graber's [LXC 1.0 blog post
......@@ -83,5 +130,3 @@ Other resources
good summary of the threats and solutions, as of August 2014
* [Linux Container Security](http://mjg59.dreamwidth.org/33170.html),
by Matthew Garrett
* [Oz](https://github.com/subgraph/oz) is a sandboxing system targeting
everyday workstation applications. By Subgraph.
......@@ -42,6 +42,14 @@ Design decisions
wrote` and will ignore tagnaq's suggestion on this; details: [doc on
reply_header_](http://kb.mozillazine.org/Reply_header_settings)
Removal of Claws-Mail
=====================
As decided in [[!tails_ticket 10010]], Claws-Mail will be [[!tails_ticket 10167 desc="removed from Tails"]]
two releases after Icedove is introduced.
Things to implement
===================
......
This reports covers the activity of Tails from February 2015 and until
the end of July 2015, as no contract was signed before August.
Everything in this report can be made public.
A small number of these deliverables are late compared to the
initially planned milestones, because without a contract and the
guarantee that they would be paid, contractors have "only" done the
vast majority of the work, but did not feel strictly bound to these
deadlines yet.
Note: the numbers preceded with a `#` correspond to tickets in our bug
tracker which contains more technical details and timeline. For example,
ticket #6938 can been seed on https://labs.riseup.net/code/issues/6938.
A. Replace Claws Mail with Icedove
==================================
Due to an important security issue that we discovered in Claws Mails [1]
we want to move faster on the migration to Icedove and try to have it in Tails
1.7 (November 3) instead of Tails 1.9 (January 26). Icedove in Tails 1.7
will fallback on the account wizard provided by TorBirdy, and Icedove in
Tails 1.9 will have the usual Icedove wizard with additional security (for
example with TLS enforced).
[1]: https://tails.boum.org/security/claws_mail_leaks_plaintext_to_imap/
B. Improve our quality assurance process
========================================
B.1. Automatically build ISO images for all the branches of our source
code that are under active development
----------------------------------------------------------------------
Any active Tails development branch is now built automatically on our
Jenkins server. Since this piece of infrastructure has been deployed,
every day at least 20 branches have been built, which exceeds the goal
(ten per day) that we had set. The latest builds can be found on
<http://nightly.tails.boum.org/>.
This is the first stage towards integrating our automated test suite
at the core of our software development process, and to run it
continuously on our development versions while they are
being developed.
The steps that lead to completion of this project where:
- B.1.1. Improve the Puppet code of our ISO building server (based on
Jenkins) to be able to replicate our setup as needed: #6938 and #6056
Once this refactoring was completed, we successfully validated it by
replicating our Jenkins setup locally. This should for example help
new contributors participate in infrastructure improvements.
- B.1.2. Have a mechanism to automatically update Jenkins jobs
when changes to their code are pushed through Git: #5898
This, along with the next few completed tasks, enabled us to have
the Jenkins build jobs dynamically generated based on activity that
happens in the main Tails code repository, in a fully automated way.
- B.1.3. Decide what kind of branches qualify for automatic building:
#8655
After an initial design was proposed by our infrastructure team,
other Tails developers were involved in the discussion to make sure
that what we were going to set up was a good match for their needs.
It was a long process, but after a few iterations we were confident
that we had identified the most important issues and resolved them,
which proved to be correct once the whole system was deployed
later on.
- B.1.4. Improve automatic cleanup of nightly built ISO images: #6439
The program that takes care of garbage collecting ISO images built
on our Jenkins infrastructure was previously buggy. Issues in the
algorithm were identified, a new algorithm was designed and
implemented, which resolved the issue. This will make it easier for
developers to debug regressions: they now have more data points
to compare.
- B.1.5. Publish and upstream our improvements to the Jenkins Puppet
module to ease its maintenance: #9682
Tails has a policy of working hand-in-hand with upstream when
feasible, as a way to share our improvements with the outside world,
and more pragmatically as a way of keeping our project sustainable
on the long run. Here, we have worked with the upstream author of
the puppet-jenkins module and had our improvements merged upstream.
As a result, we now are back to using the upstream version of
this code.
- B.1.6. Have topic branches built using the APT repository of their base branch
(stable, devel, etc.) and not devel by default as is currently the
case, merging them into their base branch. Adjust contributing
processes and documentation accordingly: #8654
We deemed it important, when building ISO images from development
branches, to make sure that they merge fine with the base branch
they should land into eventually, and in turn that they build fine
after this merge has been done. It proved to be seriously harder
than expected: we had to encode information, in our Git tree, about
which APT repositories should be used when building a Tails ISO
image from a given branch, and to adjust great parts of our
development & QA processes and infrastructure. Thankfully the result
works well and makes the daily work of Tails developers easier.
- B.1.7. Write library code that implements the automatic build
criteria and parameters designed during the previous iteration: #8657
- B.1.8. Write tools that generate a set of Jenkins jobs to build ISO
images based on the aforementioned library code: #8656
This is the implementation of the design we came up with previously
(B.1.3). The vast majority of the code was implemented as a library,
so that we can reuse it in the future for whatever may need it.
- B.1.9. Deploy, refresh regularly our Jenkins jobs
and automatically clean obsolete ones: #8658
This was the final step, when we deployed live the entire piece of
infrastructure, crossed fingers, and happily saw it work flawlessly.
B.2. Continuously run our entire test suite on all those ISO images
once they are built
-------------------------------------------------------------------
- B.2.1. Adjust our infrastructure to run tests in parallel
This was not fully completed. Some initial work has been done,
though:
* We have started adjusting our server setup, Puppet modules and
maintenance procedures to cope with multiple test runners.
* We have starting deploying more virtual machines that will act as
test runners.
- B.2.2. Decide what kind of ISO images qualify for testing
and when, how to process, advertise, and store the results
An initial draft design was written, posted on our development
mailing-list, and as a result the discussion has started but was not
completed yet.
B.3. Extend the coverage of our test suite
------------------------------------------
### Refactoring to make room for improvements in the next iterations
- B.3.1. Use libguestfs for better disk handling in the test suite:
#5294
Using libguestfs gave us some noticeable performance improvements
when running our automated test suite, and enables us more
flexibility when dealing with disk images.
- B.3.2. Reorganize the test suite: #6543
After two years of active development, our test suite has seen
a well-deserved spring clean up and refactoring, which allows us to
add more tests in a maintainable way.
### Writing more automated tests
- B.3.3. Test "Report an error" launcher: #6904
This was completed, and got rid of one more manual test we
previously had to do at release time.
- B.3.5. Write tests for I2P
Some initial work has been done, but it has not been completed yet.
- A great number of additional tests were written over the reporting
period, as part of a project we have with another sponsor.
### Make our test suite robust and faster
During this iteration, this included:
- B.3.4. Research how we can use background snapshots more efficiently
for speeding up the test suite. Report upstream any bug we discover
along the way, if any, in the hope that we can benefit more from
this feature later on: #6094
A proof-of-concept was created, and so far it is very convincing:
e.g. we've seen 20-30% test suite runtime decrease on various
hardware. We've been hit by a limitation in libvirt, and started
discussing it with upstream. Next steps are to polish our
proof-of-concept, merge it into our development branch, and report
a kernel bug to Linux upstream.
- During the reporting period, lots of efforts were put into making
our test suite more robust, as part of a project we have with
another sponsor.
C. Scale our infrastructure
===========================
C.2. Be able to detect within hours failures and malfunction on our
services
-------------------------------------------------------------------
This section is about monitoring the most critical pieces of the
infrastructure that supports Tails development. We have some early
progress to report about, even though this deliverable is technically
part of a milestone that is still many months in the future.
- C.2.3. Do an inventory of the services that need monitoring and
prioritize: #8649
We did that, and while we were at it we wrote a detailed
specification for the upcoming monitoring system, which will help
choose software and hosting.
- Some initial experiments and comparisons were made with several
monitoring systems. A dedicated virtual machine has been set up for
these experiments.
C.3. Make it easier for new contributors to start working on Tails
------------------------------------------------------------------
Until a few months ago, cloning our main Git repository required
downloading more than 300 MB of data. Most of this corresponded to
files that were not useful anymore. This made it hard for new
contributors to start working on Tails, so we had little choice but
rewriting the history of this repository. This is what we did, and as
a result our Git repository is now 75 MB large (the objective was:
less than 150 MB).
As expected, the implied migration process was a bit hard for the
least technical of Tails existing contributors (e.g. translators), but
given good documentation and support, everyone was able to go
through it.
We also made it so obsolete Git branches and APT repositories are now
identified and cleaned up regularly.
This covers deliverables C.3.1 to C.3.7.
C.4. Maintain our already existing services
-------------------------------------------
This covers deliverables C.4.1 and C.4.2.
Aside of the usual security updates and taking care of daily requests
coming from the Tails development community, here are a few selected
tasks we have completed:
* completed the migration of our main server to new hardware;
* moved the /promote/ website directory to a dedicated Git
repository so that it can grow without cluttering the main one;
* upgraded almost all our systems to Debian Jessie and adjusted our
Puppet code accordingly;
* set up an archive of the Tor Browser tarballs we need, which removes
one of the blockers so previous tags in our Git repository can still
be built in the future;
* set up a Jenkins job to proactively check for upstream merge
conflicts in our Tor Browser AppArmor profile;
* improved and distributed our backup processes;
* made it so our APT repository's keyring, and the exported APT repo's
pubkey we publish, are automatically updated;
* extended storage capacity on our main server;
* set up the infrastructure that the automated test suite team
requested (SSH, sftp, shared secrets repository);
* written a Puppet module to deploy systems able to run our
test suite.
D. Migration to Debian Jessie
=============================
We did quite some work on this migration. For details about bits that
are not covered by this OTF project, see
<https://labs.riseup.net/code/projects/tails/issues?query_id=130>.
D.2. Take advantage of systemd to improve the internals of Tails
----------------------------------------------------------------
- D.2.1. Convert at least 4 of our custom initscripts to native
systemd units
The 9 custom initscripts currently shipped in Tails (based on Debian
Wheezy) were converted to systemd unit files. Besides, while we were
at it, we converted most of our custom GNOME session initialization
scripts to systemd (9 more units in the end), as well as the MAC
address spoofing subsystem.
- D.2.2. Replace our patches to initscripts from Debian with systemd drop-in overrides
This is work in progress: our Jessie branch still patches 13
initscripts. For the record, the goal we want to reach here is to go
down to 9.
D.6. Upgrade Tails-specific tools to Debian Jessie technologies
---------------------------------------------------------------
- D.6.1 Port Tails-specific tools from udisks 1 to udisks 2: #8292
We have ported Tails Installer, Tails Upgrader, and the persistent
volume assistant to udisks 2. As a result, our Jessie branch does
not ship udisks 1 anymore, which forced us to port our automated
test suite (that still relied on it) as well.
- D.6.2 Adjust the memory wipe on shutdown feature for Debian Jessie:
#7183, #8372 and #9032
This ended up being quite involved, but was successfully completed.
Besides, we have created a Debian package (codename: wiperam) that
takes over this subsystem of Tails. The idea is that other similar
projects could benefit from it, and once all the pre-requisites have
been implemented in Debian, we could even make it available to all
Debian and Ubuntu users.
- D.6.3 Port WhisperBack, our integrated bug reporting tool, to Python 3
The initial port is done; it implied replacing some of the libraries
used by the Python 2 version, with others that support Python 3.
The last remaining step is to add native SOCKS proxy support to
this application, because for some reason the Python 3 version does
not start under torsocks: #9412.
E. Release management
=====================
- Tails 1.3 was released on 2015-02-24 [2]:
- Add Electrum bitcoin wallet.
- Sandbox Tor Browser using AppArmor.
- Add obfs4 pluggable transport.
- Add keyringer to manage and share secrets using OpenPGP and Git.
- Publish isohybrid images to simplify installation.
- Enable tap-to-click and two-finger scrolling.
- Add Ibus Vietnamese input method.
- Improve support for OpenPGP smartcards.
[2]: https://tails.boum.org/news/version_1.3/
- Tails 1.3.1 was released on 2015-03-23 [3]:
- We transitioned to a new signing key.
- Updated Tor Launcher.
[3]: https://tails.boum.org/news/version_1.3.1/
- Tails 1.4 was released on 2015-05-12 [4]:
- Update Tor Browser to include the security slider and better
protection against third-party tracking.
- Add a shortcut to gEdit in Tails OpenPGP Applet.
- Add paperkey to back up OpenPGP keys on paper.
- Better support for Vietnamese in LibreOffice.
- Disable security warnings when connecting to POP3 and IMAP.
- Add support for more printers.
- Upgrade Tor to 0.2.6.7.
- Remove the obsolete #i2p-help IRC channel.
- Remove the command line email client mutt and msmtp.
- Fix Unsafe and I2P browsers theme in Windows camouflage.
- Remove the Tor Network Settings... from the Torbutton menu.
- Upgrade syslinux for better hardware support.
- Fix the localization of Tails Upgrader.
- Fix the OpenPGP key servers configured in Seahorse.
- Prevent Tor Browser from crashing when Orca is enabled.
[4]: https://tails.boum.org/news/version_1.4/
- Tails 1.4.1 was released on 2015-07-03 [5]:
- Upgrade Tor Browser to 4.5.3.
- Upgrade Tor to 0.2.6.9-1~d70.wheezy+1+tails2.
- Upgrade Linux to 3.16.7-ckt11-1.
- Have AppArmor Deny Tor Browser access to the list of recently used files.
- Fix automatic upgrades in Windows camouflage.
[5]: https://tails.boum.org/news/version_1.4.1/
......@@ -9,25 +9,175 @@ Deliverable identifiers and descriptions are not free-form: they must
be copy'n'pasted as-is from the proposal sent to the sponsor.
</div>
<pre>
# A. Replace Claws Mail with Icedove
## A.n. description of subsection
In May we discovered than Claws Mail is leaking plaintext of encrypted
emails to the IMAP server [1] and that it was impossible to fix this
properly in Tails. As a consequence, we're now considering moving to
Icedove faster than planned initially (late January) and to:
- A.n.m. description of deliverable: ticket numbers
- Replace Claws Mail with Icedove *without* its account creation
wizard in Tails 1.7 (early November). TorBirdy already provides an
alternative account wizard which is good enough for that step.
- Add the wizard in Tails 1.9 (late January).
status summary: