Do we want to drop OATH toolkit/libs?
Currently Keysmith uses
liboath (oath-toolkit) for the OATH algorithms and Base32 decoding. However we may want to ditch this in favour of our own implementation after #6 (closed) lands. Here are a few reasons why:
- The library comes with a number of limitations.
- It is also not possible to adapt the output encoding. This turns out to be a real-world limitation: SteamGuard is basically TOTP using a custom encoding symbol table. Using the library this is impossible to support in Keysmith.
- The library comes with a creaking autotools based build system which requires at least two patches to keep working already and is generally no fun.
- In practice item 3 also means that for that reason it is currently infeasible to get Keysmith building for and working on Android (@nicolasfella)
- There is also no metadata on this library for e.g.
kdesrc-build --include-dependencies keysmithwill not work without manually installing the lib first.
- The algorithms that we need are not complex. HOTP is just an HMAC, TOTP is just a scheme to derive the counter for HOTP and so on. Qt already provides primitives (e.g.
QCryptographicHashfor hashing) and if these are unsuitable by the time #6 (closed) lands we will have integrated with 'real' crypto libs in Keysmith anyway.