Firefox Beta via Flatpak

What I've tried.

  1. Firefox beta as a snap. (Definitely easy to install. But not as quick and harder to use for managing files - makes it's own Downloads directory, etc)
  2. Firefox (stock) with custom AppArmor confinement. (Fun to do once, but the future is clearly using portals for file access, etc)
  3. Firefox beta as a Flatpak.

I've now been running Firefox as a Flatpak for over 4 months and have not had any blocking issues.

Getting it installed

Flatpak - already installed on Fedora SilverBlue (comes with Firefox with some Fedora specific optimizations) and EndlessOS at least

Follow Quick Setup. This walks you through installing the Flatpak package as well as the Flathub repo. Now you could easily install Firefox with just 'flatpak install firefox' if you want the Stable Firefox.

To get the beta you need to add the Flathub Beta repo. You can just run:

sudo flatpak remote-add flathub-beta https://flathub.org/beta-repo/flathub-beta.flatpakrepo

Then to install Firefox from it do (You can also choose to install as a user and not using sudo with the --user flag):

sudo flatpak install flathub-beta firefox

Once you run the above commend it will ask you which Firefox to install, install any dependencies, tell you the permissions it will use, and finally install.

Looking for matches…
Similar refs found for ‘firefox’ in remote ‘flathub-beta’ (system):

...posts/mindshare/snap-firefox-initial.md
   3) app/org.mozilla.firefox/x86_64/beta

Which do you want to use (0 to abort)? [0-3]: 3
Required runtime for org.mozilla.firefox/x86_64/beta (runtime/org.freedesktop.Platform/x86_64/19.08) found in remote flathub
Do you want to install it? [Y/n]: y

org.mozilla.firefox permissions:
    ipc                          network       pcsc       pulseaudio       x11       devices       file access [1]       dbus access [2]
    system dbus access [3]

    [1] xdg-download
    [2] org.a11y.Bus, org.freedesktop.FileManager1, org.freedesktop.Notifications, org.freedesktop.ScreenSaver, org.gnome.SessionManager, org.gtk.vfs.*, org.mpris.MediaPlayer2.org.mozilla.firefox
    [3] org.freedesktop.NetworkManager


        ID                                             Branch            Op            Remote                  Download
 1. [—] org.freedesktop.Platform.GL.default            19.08             i             flathub                    56.1 MB / 89.1 MB
 2. [ ] org.freedesktop.Platform.Locale                19.08             i             flathub                 < 318.3 MB (partial)
 3. [ ] org.freedesktop.Platform.openh264              2.0               i             flathub                   < 1.5 MB
 4. [ ] org.gtk.Gtk3theme.Arc-Darker                   3.22              i             flathub                 < 145.9 kB
 5. [ ] org.freedesktop.Platform                       19.08             i             flathub                 < 238.5 MB
 6. [ ] org.mozilla.firefox.Locale                     beta              i             flathub-beta             < 48.3 MB (partial)
 7. [ ] org.mozilla.firefox                            beta              i             flathub-beta             < 79.1 MB

The first 5 dependencies downloaded are required by most applications and are shared, so the actual size of Firefox is more like 130MB.

Confinement

  • You can't browsing for local files via browser file:/// (except for ~/Downloads). All local files need to be opened by Open File Dialogue which automatically adds the needed permissions. Unboxing
  • You can enable Wayland as well with 'sudo flatpak override --env=GDK_BACKEND=wayland org.mozilla.firefox (Wayland doesn't work with the NVidia driver and Gnome Shell in my setup though)

What Works?

Everything I want which includes in no particular order:

  • Netflix (some older versions had issues with DRM IMU)
  • WebGL (with my Nvidia card and proprietary driver. Flatpak installs the necessary bits to get it working based on your video card)
  • It's speedy, it starts quick as I would normally expect
  • Using the file browser for ANY file on my system. You can upload your private SSH keys if you really need to, but you need to explicitly select the file (and I'm not sure how you unshare it).
  • Opening apps directly via Firefox (aka I download a PDF and I want it to open in Evince - this does use portals for confinement).
  • Offline mode

What could use work?

  • Some flatpak commands can figure out what just "Firefox" means, while others want the full org.mozilla.firefox
  • If you want to run Firefox from the command line, you need to run it as org.mozilla.firefox. This is the same for all Flatpaks, although you can make an alias.
  • It would be more convenient if Beta releases were part of the main Flathub (or advertised more)
  • If you change your Downloads directory in Firefox, you have to update the permissions in Flatpak or else it won't allow it to work. If you do Save As.. it will work fine though.
  • The flatpak permission-* commands lets you see what permissions are defined, but resetting or removing doesn't seem to actually work.

If you think you found a Flatpak specific Mozilla bug, the first place to look is Mozilla Bug #1278719 as many bugs are reported against this one bug for tracking purposes.

Comments

Add a comment by making a Pull Request to this post.

Wrong About Signal

Updated Riot was renamed to Element. XMPP info added in comment. And Signal still doesn't let you Unregister

A couple years ago I was a part of a discussion about encrypted messaging.

  • I was in the Signal camp - we needed it to be quick and easy to setup for users to get setup. Using existing phone numbers makes it easy.
  • Others were in the Matrix camp - we need to start from scratch and make it distributed so no one organization is in control. We should definitely not tie it to phone numbers.

I was wrong.

Signal has been moving in the direction of adding PINs for some time because they realize the danger of relying on the phone number system. Signal just mandated PINs for everyone as part of that switch. Good for security? I really don't think so. They did it so you could recover some bits of "profile, settings, and who you’ve blocked".

Before PIN

If you lose your phone your profile is lost and all message data is lost too. When you get a new phone and install Signal your contacts are alerted that your Safety Number has changed - and should be re-validated.

Where profile data lives 1318.60060075387.1499999984981Where profile data livesYour Devices

After PIN

If you lost your phone you can use your PIN to recover some parts of your profile and other information. I am unsure if Safety Number still needs to be re-validated or not.

Your profile (or it's encryption key) is stored on at least 5 servers, but likely more. It's protected by secure value recovery.

There are many awesome components of this setup and it's clear that Signal wanted to make this as secure as possible. They wanted to make this a distributed setup so they don't even need to tbe only one hosting it. One of the key components is Intel's SGX which has several known attacks. I simply don't see the value in this and it means there is a new avenue of attack.

Where profile data lives 1370.275162.94704773529975250.12499999999997371.0529522647003Where profile data livesYour DevicesSignal servers

PIN Reuse

By mandating user chosen PINs, my guess is the great majority of users will reuse the PIN that encrypts their phone. Why? PINs are re-used a lot to start, but here is how the PIN deployment went for a lot of Signal users:

  1. Get notification of new message
  2. Click it to open Signal
  3. Get Mandate to set a PIN before you can read the message!

That's horrible. That means people are in a rush to set a PIN to continue communicating. And now that rushed or reused PIN is stored in the cloud.

Hard to leave

They make it easy to get connections upgraded to secure, but their system to unregister when you uninstall has been down Since June 28th at least (tried last on July22nd). Without that, when you uninstall Signal it means:

  • you might be texting someone and they respond back but you never receive the messages because they only go to Signal
  • if someone you know joins Signal their messages will be automatically upgraded to Signal messages which you will never receive

Conclusion

In summary, Signal got people to hastily create or reuse PINs for minimal disclosed security benefits. There is a possibility that the push for mandatory cloud based PINS despite all of the pushback is that Signal knows of active attacks that these PINs would protect against. It likely would be related to using phone numbers.

I'm trying out the Element which uses the open Matrix network. I'm not actively encouraging others to join me, but just exploring the communities that exist there. It's already more featureful and supports more platforms than Signal ever did.

Maybe I missed something? Feel free to make a PR to add comments

Comments

kousu posted

In the XMPP world, Conversastions has been leading the charge to modernize XMPP, with an index of popular public groups (jabber.network) and a server validator. XMPP is mobile-battery friendly, and supports server-side logs wrapped in strong, multi-device encryption (in contrast to Signal, your keys never leave your devices!). Video calling even works now. It can interact with IRC and Riot (though the Riot bridge is less developed). There is a beautiful Windows client, a beautiful Linux client and a beautiful terminal client, two good Android clients, a beautiful web client which even supports video calling (and two others). It is easy to get an account from one of the many servers indexed here or here, or by looking through libreho.st. You can also set up your own with a little bit of reading. Snikket is building a one-click Slack-like personal-group server, with file-sharing, welcome channels and shared contacts, or you can integrate it with NextCloud. XMPP has solved a lot of problems over its long history, and might just outlast all the centralized services.

Bryan Reply

I totally forgot about XMPP, thanks for sharing!

Don't Download Zoom!

First, I strongly recommend switching to Jitsi Meet:

  • It's free
  • It doesn't require you to sign up at all
  • It's open source
  • It's on the cutting edge of privacy and security features

Second, Anything else that runs in a browser instead of trying to get you to download an specific desktop application. Your browser protects you from many stupid things a company may try to do. Installing their app means you are at more risk. (Apps for phones is a different story.).

A small sampling of other web based options:

  • Talky.io (also open source, no account required)
  • 8x8.vc which is the company that sponsors Jitsi Meet. Their offering has more business options
  • Whatever Google calls their video chat product this week (Duo, Hangouts, Meet).
  • join.me
  • Microsoft Skype (no signups or account required for a basic meeting!)
  • whereby

There are many reasons not to choose Zoom.

😞😞😞

Finally, So you have to use Zoom?

Zoom actually supports joining a call with a web browser. They just don't promote it. Some things may not work as well but you get to keep more of your privacy and security.

  1. On joining the meeting close the request to run a local app.
  2. Click Launch Meeting in middle of screen. Zoom join meeting page
  3. Again close out of the request to open a local app
  4. Ideally, you now get a join from browser, click that! Click join from browser

If it doesn't work try loading the site in another browser. First try Chrome (or those based on it - Brave/Opera) and then Firefox. It's possible that your organization may have disabled the join from web feature.

If you are a Zoom host or admin (why?) you can also ensure that the web feature is not disabled.

2020 LiveCD Memory Usage Compare

Time for a 20.04 LTS LiveCD memory comparison with a bunch more distros. I last did one in 2016.

Using Lubuntu as an example base memory usage approximately doubled from 2016 (251M) to 2020 (585M). Those numbers aren't strictly comparable because I'm not using the exact same setup as in 16.04 and I enabled more modern features (virtio graphics, EUFI, 4 cores).

Memory usage compared (in G) 000.20.20.40.40.60.60.80.8111.21.21.41.41.61.61.81.8222.22.2Clear 33300Elementary 5.1Endless 3.8Fedora 32KubuntuLubuntuManjaro 20.0.3 XFCEopenSUSE Leap 15.1Solus 4.1UbuntuUbuntu BudgieUbuntu MateXubuntu0.822.259000436946966356.3326446313486Clear 333000.869.95685851611904356.3326446313486Elementary 5.11117.6547165952911337.5288111415677Endless 3.81.25165.35257467446323314.02401927934153Fedora 320.8213.05043275363525356.3326446313486Kubuntu0.585260.7482908328073376.54676563286307Lubuntu0.9308.44614891197944346.93072788645816Manjaro 20.0.3 XFCE1.25356.14400699115146314.02401927934153openSUSE Leap 15.11403.84186507032354337.5288111415677Solus 4.11451.53972314949556337.5288111415677Ubuntu1499.2375812286677337.5288111415677Ubuntu Budgie0.9546.9354393078397346.93072788645816Ubuntu Mate0.6594.6332973870118375.1364781211295Xubuntu1.536.250372140170775290.51922741711536Clear 333001.2583.94823021934286314.02401927934153Elementary 5.11.5131.6460882985149290.51922741711536Endless 3.81.5179.34394637768705290.51922741711536Fedora 321.25227.04180445685907314.02401927934153Kubuntu0.7274.7396625360311365.73456137623907Lubuntu1.5322.43752061520325290.51922741711536Manjaro 20.0.3 XFCE1.75370.1353786943753267.0144355548892openSUSE Leap 15.11.5417.83323677354736290.51922741711536Solus 4.11.5465.5310948527194290.51922741711536Ubuntu1.5513.2289529318915290.51922741711536Ubuntu Budgie1.25560.9268110110635314.02401927934153Ubuntu Mate1.25608.6246690902356314.02401927934153Xubuntu1.7550.24174384339459267.0144355548892Clear 333001.7597.93960192256667267.0144355548892Elementary 5.11.75145.63746000173873267.0144355548892Endless 3.81.75193.33531808091084267.0144355548892Fedora 321.75241.03317616008286267.0144355548892Kubuntu0.9288.73103423925494346.93072788645816Lubuntu1.75336.4288923184271267.0144355548892Manjaro 20.0.3 XFCE2384.1267503975991243.50964369266302openSUSE Leap 15.11.75431.8246084767712267.0144355548892Solus 4.12.25479.5224665559432220.00485183043688Ubuntu2527.2203246351154243.50964369266302Ubuntu Budgie1.75574.9181827142874267.0144355548892Ubuntu Mate1.5622.6160407934594290.51922741711536XubuntuMemory usage compared (in G)Boots to DE that can start somethingBrowser load simple websiteYouTube plays Big Buck Bunny maximi…YouTube plays Big Buck Bunny maximized

Lubuntu is able to work with less at least partially because of Zram. The other distro that has Zram enabled is Endless, but they also use the Chromium browser which generally uses more memory than Firefox (also Elementary uses Ephipany). My guess is if Xubuntu enabled zram it's profile would more closely match Lubuntu.

Notes:

  • Time limit for each applicaton launch is approximately 30 seconds.
  • Accuracy over 1G is by .25G increments. Under 1G, I tried to narrow it down to at least .1G.
  • Getting out of full screen on YouTube apparently is an intensive task. Dropped testing that.
  • Screen size was set to 1080p/60Hz.
  • Sample qemu line: qemu-system-x86_64 -enable-kvm -cdrom clear-33300-live-desktop.iso -smbios file=/usr/share/ovmf/OVMF.fd -m 1024M -smp 4 -cpu host -vga virtio --full-screen
  • All Ubuntu derivatives were from 20.04 LTS

Quick Rust Comparison

I've been wanting to try out Rust with something very simple as a first pass through the language.

Rust Impressions

Although I didn't do much with functions on this quick pass - I love the ability to not have the order of main in a program to matter.

Super helpful error messages. Here is an example:

warning: value assigned to `temp` is never read
 --> src/main.rs:4:13
  |
4 |     let mut temp=0u32;
  |             ^^^^
  |
  = note: `#[warn(unused_assignments)]` on by default
  = help: maybe it is overwritten before being read?

I know others have said this, but the Rust compiler feels like it was designed to help me code, rather than just throw errors.

Speed?

I decided to write a simple unoptimized version of the fibonacci sequence. My goal was to take enough time to be noticable...

On my first pass:
  • Rust runs took 1m34seconds (using cargo run)
  • Python took more than 6 minutes
  • C got 7 seconds

Clearly I must have done something wrong...

It turns out that by default it has debug info and checks that slow Rust down. So a

cargo build --release
./target/release/fib

Then it was faster than C.. and I realized I need to turn off C\'s debug bits too with:

gcc -O2 -s -DNDEBUG to gcc helped. gcc  fib.c
The final results (all approximate):
  • Python: 6+ minutes.
  • C: 1.101s
  • Rust: .95sE

The Rust

fn main() {
    let mut previous=0u32;
    let mut current=1u32;
    let mut temp;
    let maxvalue = 2000000000u32;

    for _n in 0..2000000000 {
            if current >= maxvalue {
                //Reset!
                previous=0; current=1;
            }
        temp = current;
        current = previous + current;
        previous = temp;
    }
    println!("{}", current);
}

The C

#include <stdio.h>
int main() {

    unsigned long int previous=0;
    unsigned long int current=1;
    unsigned long int temp;
    unsigned long int maxvalue = 2000000000;
    for ( int n=0; n < 2000000000; n++ ) {
        if (current >= maxvalue) {
                //Reset!
                previous=0; current=1;
        }
        temp = current;
        current = previous + current;
        previous = temp;
    }
    printf("%lu", current);
}

The Python3

previous=0;
current=1;
temp = 0;
maxvalue = 2000000000;

for n in range(2000000000):
    if current >= maxvalue:
        #Reset!
        previous=0; current=1;
    temp = current;
    current = previous + current;
    previous = temp;
print(current);

3 Malaysia MPEG-2 Patents left

With February 13th passing it would appear there are only 3 Malaysia patents left:

  • MY 128994 (possible expiration of 30 Mar 2022)
  • MY 141626-A (possible expiration of 31 May 2025)
  • MY-163465-A (possible expiration of 15 Sep 2032)

These two just expired:

  • MY 118734-A - Exp. Jan 31, 2020
  • PH 1-1995-50216 - Exp. Feb 13, 2020

I am very much not a patent lawyer, but my reading indicates all the 3 remaining are really all the same expired US Patent US5565923A with varying Grant dates causing to expire far in the future.

I've started a detailed tracker for those who want more details.

Hack Computer review

I bought a hack computer for $299 - it's designed for teaching 8+ year olds programming. That's not my intended use case, but I wanted to support a Linux pre-installed vendor with my purchase (I bought an OLPC back in the day in the buy-one give-one program).

I only use a laptop for company events, which are usually 2-4 weeks a year. Otherwise, I use my desktop. I would have bought a machine with Ubuntu pre-installed if I was looking for more of a daily driver.

The underlying specs of the ASUS Laptop E406MA they sell are:

Unboxing and first boot

Unboxing

Included was an:

  • introduction letter to parents
  • tips (more for kids)
  • 2 pages of hack stickers
  • 2 hack pins
  • ASUS manual bits
  • A USB to Ethernet adapter
  • and the laptop:

Laptop in sleeveLaptop out of sleevefirst open

First boot takes about 20 seconds. And you are then dropped into what I'm pretty sure is GNOME Initial Setup. They also ask on Wifi connections if they are metered or not.

first open

There are standard philips head screws on the bottom of the laptop, but it wasn't easy to remove the bottom and I didn't want to push - I've been told there is nothing user replaceable within.

The BIOS

The options I'd like change are there, and updating the BIOS was easy enough from the BIOS (although no LVFS support..).

bios ez modebios advanced

A kids take

Keep in mind this review is done by 6 year old, while the laptop is designed for an 8+ year old.

He liked playing the art game and ball game. The ball game is an intro to the hack content. The art game is just Krita - see the artwork below. First load needed some help, but got the hang of the symmetrical tool.

He was able to install an informational program about Football by himself, though he was hoping it was a game to play.

AAAAAmy favoritewater color

Overall

For target market: It's really the perfect first laptop (if you want to buy new) with what I would generally consider the right trade-offs. Given Endless OS's ability to have great content pre-installed, I may have tried to go for a 128 GB drive. Endless OS is setup to use zram which will minimize RAM issues as much as possible. The core paths are designed for kids, but some applications are definitely not. It will be automatically updating and improving over time. I can't evaluate the actual Hack content whose first year is free, but after that will be $10 a month.

For people who want a cheap Linux pre-installed laptop: I don't think you can do better than this for $299.

Pros:

  • CPU really seems to be the best in this price range. A real Intel quad-core, but is cheap enough to have missed some of the vulnerabilities that have plagued Intel (no HT).
  • Battery life is great
  • A 1080p screen

Cons:

  • RAM and disk sizes. Slow eMMC disk. Not upgrade-able.
  • Fingerprint reader doesn't work today (and that's not part of their goal with the machine, it defaults to no password)
  • For free software purists, Trisquel didn't have working wireless or trackpad. The included USB->Ethernet worked though.
  • Mouse can lack sensitivity at times
  • Ubuntu: I have had Wifi issues after suspend, but stopping and starting Wifi fixed them
  • Ubuntu: Boot times are slower than Endless
  • Ubuntu: Suspend sometimes loses the ability to play sound (gets stuck on headphones)

I do plan on investigating the issues above and see if I can fix any of them.

Using Ubuntu?

My recommendations:

  • Purge rsyslog (may speed up boot time and reduces unnecessary writes)
  • For this class of machine, I'd go deb only (remove snaps) and manual updating
  • Install zram-config
  • I'm currently running with Wayland and Chromium
  • If you don't want to use stock Ubuntu, I'd recommend Lubuntu.

Dive deeper

Stop changing the clocks

Florida, Tennessee, the EU and more are considering one timezone for the entire year - no more changing the clocks. Massachusetts had a group study the issue and recommend making the switch, but only if a majority of Northeast states decide to join them. I would like to see the NJ legislature vote to join them.

Interaction between countries would be helped by having one less factor that can impact collaboration. Below are two examples of ways this will help.

Meeting Times

Let's consider a meeting scheduled in EST with partipants from NJ, the EU, and Arizona.
NJ - normal disruption of changing times, but the clock time for the meeting stays the same.
Arizona - The clock time for the meeting changes twice a year.
EU - because they also change their clocks at different points throughout the year. Due to this, they have 4 clock time changes during each year.

This gets more complicated as we add partipants from more countries. UTC can help, but any location that has a time change has to be considered for both of it's timezones.

Global shift work or On-call

Generally, these are scheduled in UTC, but the shifts people actually work are in their local time. That can be disruptive in other ways, like finding child care.

In conclusion, while these may be minor compared to other concerns (like the potential health effects associated with change the clocks), the concerns of global collaboration should also be considered.

Now powered by GitLab, Nikola, and Cloudlfare

I just finished moving my website from Wordpress to Nikola (static site generator), GitLab (git and hosting), and CloudFlare (CDN, HTTPS and more).

Why Nikola

Their attitude in the handbook is "DON'T READ THIS MANUAL. IF YOU NEED TO READ IT I FAILED, JUST USE THE THING." That's my kind of software methodology. Don't blame the user, make the system better.

It is also a great handbook that has had pretty much every question I've asked. Documentation is still essential, but it's nice if the commands are self explanatory.

It just worked to import my Wordpress site (minus comments which I "inlined" or deleted for various reasons). I did do some manual HTML to markdown conversion for pages I want to edit more.

Why GitLab

I first tried and had Nikola working with GitHub, but GitLab gives me:

  • Automatic building - I don't have to have a separate branch for output, I just git push my changes (or change on the website) - and GitLab will run a job to create my website. I know this is possible on GitHub, but GitLab just makes it easy.
  • The option to upload SSL Certs. If I need to drop CloudFlare for some reason, I can have GitLab maintain my website using HTTPS (Which I need to because I'm on the HSTS preload list).
  • Easier drive by contributions. GitLab lets you sign in with Google, Twitter, GitHub, or BitBucket. I'm thinking for suggesting changes to say a paper (or even this blog post!), that will make for a lower barrier to entry. (Of course, I'd prefer any OpenID but it's better than requiring a new account)

I absolutely love that they have their company handbook maintained in Git and public to the world (with merge request welcome!).

Why CloudFlare

CloudFlare's free plan rocks. And if I ever need to be able to handle more traffic faster, I can upgrade/downgrade as necessary.

  • Free (Good) SSL with TLS1.3 Beta too
  • Free IPv6
  • Free HTTP2

Who we trust | Building a computer

I thought I was being smart.  By not buying through AVADirect I wasn't going to be using an insecure site to purchase my new computer.

For the curious I ended purchasing through eBay (A rating) and Newegg (A rating) a new Ryzen (very nice chip!) based machine that I assembled myself.   Computer is working mostly ok, but has some stability issues.   A Bios update comes out on the MSI website promising some stability fixes so I decide to apply it.

The page that links to the download is HTTPS, but the actual download itself is not. I flash the BIOS and now appear to have a brick.

As part of troubleshooting I find that the MSI website has bad HTTPS security, the worst page being:

Given the poor security and now wanting a motherboard with a more reliable BIOS  (currently I need to send the board back at my expense for an RMA) I looked at other Micro ATX motherboards starting with a Gigabyte which has even less pages using any HTTPS and the ones that do are even worse:

Unfortunately a survey of motherboard vendors indicates MSI failing with Fs might put them in second place.   Most just have everything in the clear, including passwords.   ASUS clearly leads the pack, but no one protects the actual firmware/drivers you download from them.

Main Website Support Site RMA Process Forum Download Site Actual Download
MSI F F F F F Plain Text
AsRock Plain text Email Email Plain text Plain Text Plain Text
Gigabyte (login site is F) Plain text Plain Text Plain Text Plain text Plain Text Plain Text
EVGA Plain text default/A- Plain text Plain text A Plain Text Plain Text
ASUS A- A- B Plain text default/A A- Plain Text
BIOSTAR Plain text Plain text Plain text n/a? Plain Text Plain Text
A quick glance indicates that vendors that make full systems use more security (ASUS and MSI being examples of system builders).

We rely on the security of these vendors for most self-built PCs.  We should demand HTTPS by default across the board.   It's 2017 and a BIOS file is 8MB, cost hasn't been a factor for years.