In the latest commit QR_based_applet/build.xml has the following line altered (line 11):

from:
<cap targetsdk="../oracle_javacard_sdks/jc221_kit" jckit="../oracle_javacard_sdks/jc305u3_kit" output="bin/QRSTK.cap" sources="src" classes="bin" version="1.0">

to:

The missing ../ causes the build to fail.

@hkscy is there a specific reason for this (do you have a different local directory structure?), or can we revert?

when running QR_based_applet/host/setup_fs.py

You get an import error like this for newer version of Pysim:

ImportError: cannot import name 'card_handler' from 'pySim.card_handler'

You need to update the host/setup_fs.py file to conform to the current version of pySim. Change two of the import statements as follows (note the inconsistent use of camelcase and underscores):

from pySim.card_handler import CardHandler
from pySim.cards import card_detect, SimCard

and change the use of the (old) card_handler class on line 120:

cardhandler = CardHandler(sl)

After chatting with @lukehare, it is apparent that the repo structure is confusing. Below are some notes for cleaning up the repo and ways of working. It is worth spending time on this now because it will only get more cluttered as we go.

• Ways of working: start using issue tracking more readily, have a develop branch and branch naming fix/XX, feature/XX where <XX> is issue number.
• Delete unused/commented code. There are also a bunch of development scripts and it is unclear what is necessary and what is development.
• Update head README in line with Tim's installation notes.
• There are at least two setup_fs.py files.
• redundant .gitignore and .gitmodules files
• we have two host directory. Consider renaming for clarity.
• in a few places you need to copy and paste simcard related stuff like ADM1 keys. Have a file where you can keep these (and add them to .gitignore) and have the codebase pick them up automatically.
• should sim-tools be a forked repo then added submodule, rather than a folder?
• can we have script (e.g. makefile) that automates much of the installation?

Please add to this list.

There are two virtual environments needed, one for pySim that uses python3, and one for stkp2.

Confusingly you need to use a different environment even in the same subfolder, dependent on the argument (see notes).

I suggest that we use Poetry or pyenv to manage environments, and refactor the repo so that it is clearer what parts of the code need which environments.

As discussed 8/11/21 meeting. There are some errors and omissions owing to earlier merging of two code bases.

As an alternative to displaying an icon using the STK, it may be possible to use the STK to proactively display an MMS containing a QR code image. This may offer improved image resolutions and flexibility of approach or device support.

We don't need to implement all the features. Just the menus. This will lower significantly the cost of using our solution.

In the current workflow you cannot test changes in code until you load the sim into a phone. This is inefficient.

We are attempting to create a dev environment using Eclipse + Javacard plugin.

Ongoing notes here

Which method did you use for this? When I navigate the Sysmocom cards using pysim the DF.Graphics directory is absent.

QREncoder takes about 15s.

It would be helpful to optimise this, and we can do it right away (i.e. it doesn't need the Javacard environment).

Implement crypto algorithms on the SIM.

These are:

• HOTP
• ECDSA signing with key derivation on the instance key
• BSI standard domain-unsinkable sigtnatures

e.g., replicate hosts/convert_image.py using Java Card API

Stock Android and KaiOS do not fully support the display of full-screen icons or images loaded from the SIM card. Android (or FirefoxOS as a proxy for KaiOS compatibility) could be modified to make this possible!

In a bit more detail, the STK implementation for Android can be found at e.g., https://android.googlesource.com/platform/packages/apps/Stk/+/eclair-release/src/com/android/stk/StkDialogActivity.java

If you take a look at e.g., line 95-96, you can see that icon are implemented as:

window.setFeatureDrawable(Window.FEATURE_LEFT_ICON, new BitmapDrawable(mTextMsg.icon));

which means that DISPLAY_TEXT icons are only ever displayed using the Android Window.FEATURE_LEFT_ICON element. This limits their size and usability.

Some specific (unordered and overlapping) tasks are:

• Modify android or KaiOS DISPLAY_TEXT so that if the icon_replaces_test flag is used in the STK, the icon is drawn 'full screen' or at least largely
• Create an Android app to display the icon from the SIM, using a different mechanism than the STK
• Investigate the Android STK to see if there any other ways of displaying an icon that are supported or which we can exploit. DISPLAY_TEXT is not the only command we have access to! (e.g. see play_tone and how that displays the icon differently - https://android.googlesource.com/platform/packages/apps/Stk/+/eclair-release/src/com/android/stk/ToneDialog.java)

We would like to set up a Javacard development environment on an Azure VM that we all have access to in order to streamline the development process (currently, this is quite slow as it requires re-building and installing the applet on SIM in order to test, and gives uninformative error messages in case of bugs).

This HackMD contains the steps we have taken so far:
https://hackmd.io/CW_Zw2T1Tje0rOQE4OpaKw?both

(To be populated)

Get Icons to display as BIG as possible. Current demo only shows a 27x27 icon.

The directory structure should reflect how the repository is intended to be used. And a logical directory structure will help separate, for example, things that need a different environment, are written in a different language, are used for a different purpose, or require a different suite of tests etc.

It is sometimes difficult to come up with a perfectly logical directory structure ahead of time because this is a research project. But it is worth thinking about how we see the code developing.

It seems to me and @lukehare that there are three parts to the repo:

• git submodules (sim-tools, oracle_javacard_sdks)
• applets (which is actually the toolbox).
• utils (this is the root host folder - better name required as I'm not sure how the scripts in this folder are related).

Here are some example directory structures that would be useful and extensible, depending on whether the QR and text applets are ultimately intended to be the same app.

Let's discuss, @hkscy, @mavroudisv, @thobson88

A)

- submodules
    - oracle_**
    - sim-tools ** (fork)
- toolbox
    - QR
        - install
            - poetry.lock / pyenv.version (for easily managing different python requirements)
        - development
            - poetry.lock / pyenv.version
        - applet/java
            - lib
            - src
            - build.xml
        
        - README
    - text
        - install
        - development
        - applet/java
            - lib
            - src
            - build.xml
            - README
    - etc.
    - README
- root-host folder [better name!, utils?]
- highlevel README + installation
- .gitignore
- .gitmodules

B)

- submodules
    - oracle_**
    - sim-tools ** (fork)
- toolbox
    - applet
        - install
        - development
            - QR
            - text
        - java
            - lib
            - src
            - build.xml
    - etc.?
    - README
    
- root-host folder [better name!, utils?]
- highlevel README + installation
- .gitignore
- .gitmodules

On slack. There is a video showing a weird bug using commit 8f2cfebef78822ecf23cc9c938d2bf3e7d42b90b.

With limited knowledge of the underlying code: the icon buffer doesn't clear and icons accumulate. And bigger icons do not display well.