These are chat archives for jdubray/sam

2nd
May 2016
Jean-Jacques Dubray
@jdubray
May 02 2016 14:15
@gfortaine thank you Guillaume, looks very interesting indeed. Any chance we could share the documentation of the APIs for the TODO example: http://todolist.kuzzle.io/?
For those of you interested, reactjs/redux#1528 have engaged in a discussion. From my perspective we are making good progress in establishing some shared understanding of the similarities and differences between underlying principles of Redux and SAM. My key take aways at this point of the discussion:
  • following Erik Meijer advice: embrace impurity => no reason for the reducer to be pure
  • thunks are pretty much (smart) wiring, you can use functions just as well as actions and follow a propose/accept approach
  • I need to think a bit more about Sagas, but I don't think it's possible (or even necessary) to isolate side-effects
devin ivy
@devinivy
May 02 2016 14:40
how does that last point affect time-travel?
Jean-Jacques Dubray
@jdubray
May 02 2016 14:42
The fact and the matter is that the Redux community cannot deliver (yet) a working Time Travel tool as soon as there are side-effects some where. IMHO, I don't think this is possible to make it work even if you could isolate them.
It seems logical to me, as soon as the system is non deterministic, you can't make it deterministic by sheer factoring (i.e. isolating the non deterministic parts). I don't see a way around it.
devin ivy
@devinivy
May 02 2016 14:54
and the SAM time-traveling is similarly ignoring side-effects?
the countdown example
Jean-Jacques Dubray
@jdubray
May 02 2016 14:57
well, it will exhibit the same problem, otherwise time travel would be possible for humans too since the universe is just a giant state machine. the only reason you cannot time travel, is because you can never go back to a particular state.
devin ivy
@devinivy
May 02 2016 14:57
so, side-effects are our app's entropy :P
Jean-Jacques Dubray
@jdubray
May 02 2016 15:00
entropy is a bit different, I am really talking about state machines. TT works when you control 100% of the State.
Brian Haak
@avesus
May 02 2016 17:44
@jdubray Just add me on Facebook and Twitter to have more instant conversation. I've invented a part of the theory of everything along with ideas of artificial general intelligence and super-compression algorithms.
Jean-Jacques Dubray
@jdubray
May 02 2016 17:45
You are both based in Seattle, we can even meet in person if you'd like!
Brian Haak
@avesus
May 02 2016 17:45
I see you thinking in the same direction. You and Yelouafi inspired me a lot. Recommend you to read the works of Gordon Pask.
Awesome. I'll find my time in June.
Jean-Jacques Dubray
@jdubray
May 02 2016 17:47
ok
devin ivy
@devinivy
May 02 2016 17:48

how does this relate to

the theory of everything along with ideas of artificial general intelligence and super-compression algorithms

?

Brian Haak
@avesus
May 02 2016 18:23
Ok, I'll explain.
Is it possible to do in a private chat because it has a long of words. If you're interested. ;-)
Brian Haak
@avesus
May 02 2016 18:55
In two words, by the virtue of any intelligence and consciousness and the sameness of the world's structure. The best compression algorithm is automatically the state machine's code of AGI. Understating of AGI is at the same time understanding of the Universe.
brusand
@brusand
May 02 2016 20:03
turing le génie : extract " A key feature of Turing machines is that both the state of the machine and the current symbol on the tape being read in, are necessary to determine the future evolution of the system. " Turing anf information theory based of origin of life http://rsif.royalsocietypublishing.org/content/10/79/20120869.full
brusand
@brusand
May 02 2016 20:10
this appears to be very similar to the case presented by biological systems where the update rules change in response to information read-out from the current state (as we discuss below, both are an example of top-down causation via information control).
on tutoie l universalité ici lolllll
Brian Haak
@avesus
May 02 2016 20:13
Yes. And to understand all the things just remind Einstein and think in 4D. Time is just a dimension.
Jean-Jacques Dubray
@jdubray
May 02 2016 20:13
We just need someone to connect TLA+ to Quantum Electrodynamics... wait...
Brian Haak
@avesus
May 02 2016 20:39
Yelouafi has only actions and effects. And state. Actually, no pure state machines there. But he invented the genius way to scale things well. To compose and componentize. But TLA+ opens my mind and I mixed all that things together with my own ideas on actor model and linked networks state representation. We live in awesome times, guys.
I think we soon find the guy who knows QED pretty well. Hey, people, here soon will be created true Artificial General Intelligence and Theory of Everything ;-)
But we have to work openly together and talk with each other.
Stardrive Engineering
@HighOnDrive
May 02 2016 21:10
"Actually, no pure state machines there. But he invented the genius way to scale things well. To compose and componentize." :+1:
Stardrive Engineering
@HighOnDrive
May 02 2016 21:16
Don't want to end up being a religion like CycleJS, chastising every side-effect and getting hopelessly bogged down into creating drivers for everything under the sun. There are simply better things to do like developing better reactive loop architectures that compose well and flow naturally.
Jean-Jacques Dubray
@jdubray
May 02 2016 21:27

+1 - that's exactly the nature of the discussion I have with some people in the Redux community. I believe the question is on the table:

  • why Redux' actions are really "events" (relay something that has happened)
  • why does the reducer has to be a pure function
  • why try to coral side-effects in a single place?

IMHO, actions are functions, we need to embrace impurity and we should implement side-effects where they make sense.

If we come up with this kind of programming model, a series of wiring options from basic to advanced and several architecture options ranging from serverless (AWS Lambda) to browser only, we may well collectively revolutionize app development.
Stardrive Engineering
@HighOnDrive
May 02 2016 21:43
Sounds good, breaking out the reactive flow is more important than ever! Because this has been largely missing frameworks proposing entangled short sighted solutions are being over rated and misleading many down a path of dead ends.
Jean-Jacques Dubray
@jdubray
May 02 2016 21:45
That's a pretty strong statement coming from you!
Stardrive Engineering
@HighOnDrive
May 02 2016 21:45
I'm following the Redux discussion, most interesting: reactjs/redux#1528
Well, I'm always hoping, yet have to keep moving along :smile:
Jean-Jacques Dubray
@jdubray
May 02 2016 21:47
Yes, I think we finally got some good Q&A going. I think that the "action cancellation" is perhaps what was the ice breaker.
Stardrive Engineering
@HighOnDrive
May 02 2016 22:16
I was pleasantly surprised to see that dataflow was being composed/assembled in a way that has leanings toward my own efforts, see tweet :sparkles: