Skip to content
Mark Nadal edited this page Jul 14, 2018 · 3 revisions

Question

@amark Should this test fail or succeed?:

  it('should accept "incoming" when null is "current"', () => {
    const state = 2;
    expect(HAM(state, state, state, 'a', null as any)).toEqual({ converge: true, incoming: true });
  });

Answer

@Kuirak_twitter if null as any is null then it should fail :( the reason why is I at some point in time wanted to define my own type rules, that null, false, true, etc. as types with lower atomic values than strings, therefore... naturally upon a direct state conflict (2 === 2) the "higher" value was deterministically chosen (aka 'a' is incoming, and is > than null therefore it wins). However... as you might expect... :/ this caused all sorts of non-standard type priorities between different programming languages and parsers I checked :( :( :( So I unfortunately had to make the more (currently) universal decision: Let the serializer decide, as various programming languages are forced to be compatible with serialization protocols. Aka, JSON :( (which is terrible terrible terrible), I do have my own serialization format but it isn't nearly as adopted as JSON.

so.... as you can guess, what happens? JSON.stringify('a') < JSON.stringify(null) :( means null has higher precedence. :(

(for those who don't know... the reason why JSON does this is that strings are encoded with " which is ASCII code 34 [one up from RAD's starting key ! bang! at 33, the lowest ASCII visible character. (whether " " is visible is debatable or not, either way, the reason why I didn't choose RAD to start at that is because filesystems don't like " ")] and JSON encodes null as the serialized text of null without the parse type indicator of " since it isn't a string [ASCII encodes null at 0 which makes sense, but most systems/parsers won't transmit (or improperly handle) ASCII 28 and below]. And in serialized format, "(a") < n(ull). :( :( :(. )

Note: This also applies to false < null < true!


Oh, one last thing: As a mathematician I care about the meaningfulness of an operation (I'd think null < "a") but when it comes to the goal of GUN (deterministic convergence) the most important thing is correctness not meaning because meaning can always be built/constructed ontop of GUN (like with SEA, it rejects data), and the reality is that if 2 users at the exact same micro-millisecond do conflicting null vs 'a' the appropriate and meaningful response is for those users to react to the change and then (a few seconds later) change the data to what they intended. The more the machine (the conflict resolution algorithm) can not impose its own "meaning" (even if it is coming from as elegant of a mathematician as me ;) who is provably right ;) hehe) the better, or else the harder and harder the system becomes to reason about (even if the deterministic choices it makes don't make intuitive sense, at least it is computationally deterministically guaranteed... even if by arbitrary serializations standards committees).

This wiki is where all the GUN website documentation comes from.

You can read it here or on the website, but the website has some special features like rendering some markdown extensions to create interactive coding tutorials.

Please feel free to improve the docs itself, we need contributions!

Clone this wiki locally