futureofcoding.org

Jonathan Edwards 6/14/18 11am ET

• Never heard of PPIG. What conferences do you go to?
  • <programming>
  • Strangeloop
  • Splash LIVE
  • Talk vs attending
  • What value do you get?
  • Do you schedule meetings?
• Methodology problem (or is it a collaboration problem?)
  • Coolness isn’t right criteria
  • What about success of a project or methodology?
  • Is the goal “getting ideas out there” (BV or Alan Kay) or transferring these ideas into a project (Haskell or Google)?
• My current research direction
  • Ultimate goal: truly open source software, where people can fork/customize their applications in real time. A world where 10% (instead of 0.01%) of users can edit their tools, would be a world with lots of amazing, free tools.
    • Sub-goal: people be able to navigate and understand the relevant section of the project they are trying to modify, and make the modification
      • Any 10x or 100x programmer has trouble with this.
        • Files and folders. Ugh. What runs when?
        • Imperative programming: need to read every line
      • Solution:
        • Remove all mutation from afar (explicit dependencies)
        • Remove control flow
        • Either FRP or synchronous languages
          • Then we’ll have a declarative specification that we can build visual metaphor abstractions on top of
    • Other things we’ll need that I’m not thinking about
      • A version control system that could handle many, many divergent forks
      • A way to share data between applications - a more semantic format, maybe
      • The visual metaphor layer
      • The “editing” layer that lives in each application
      • The hosting and permissioning, and accounts for each application
  • Current challenges
    • Explaining to my friends and family what I do - which isn’t hard when things are going well
    • Making enough money, so I can live, but also have enough time for this
    • Picking which problems to work on, and how they fit into a larger solution/vision
    • Collaboration with others
      • You’ve worked at MIT and YC research
      • Haskell is such an interesting success story of collaboration around laziness
    • Reading papers vs coming up with my own ideas
    • Presenting work vs working on new ideas
    • Conferences?
• Notes
  • Don’t read wadler
  • https://arxiv.org/abs/1803.10195
  • Notes on my direction
    • Who hits wants
    • It would be nice to point to a piece of code:
      • what parts of the state does it depend on and what does it modify
    • Subsetting the project enough so I can make some progress on it
      • Pick a language to work within
    • A issue we’re torn on:
      • Nulls are formally messy but informally flexible and natural in some sense
      • Andrew Myers from CMU HCI thinks nulls are natural
    • Generally strategy:
      • Yes tree-structured state
        • We structure our UIs and code as trees already
        • It’s easy to map dbs as trees (and no sql already is)
        • We can unify all our data into one structure
        • Tree data model programming language https://en.wikipedia.org/wiki/ParaSail_(programming_language)
      • No arbitrary graphs which can go any which way are too complicated: problems of pointers and heaps

Direct programming Notes

• Presentational
  • I wouldn’t start by opening the base field. A better time may be when you should how you can drill down to any definition.
  • Don’t need to explain the cursor location, action and optionally an argument
  • The guiding principle: start with that, before explaining how the document works?
    • Start with the goal of minimizing learning curve of beginner, then talk about how indirection is hard for them
  • Sorry for getting the event invite in your video!
  • How to evaluate the work
    • See if it solves the indirection problem by teaching it to some students…?
• Meat
  • I wouldn’t call it “Past State: date”. Maybe just the date. I prefer the word “version” to “state” for end user programmers.
  • I’d highlight the history and state with different colors. One should be a “selection color” and the other should be a “change was made here color”
  • I think the no-op default is exactly what you don’t want. You want it to change anything, so I’d default them all to 2.
  • I don’t like the right-click multiply. I think I’d much, much prefer having to type 3 * 5 in the a-number field, and then on enter or leaving the field, having it evaluate and then add the multiplication to the history.
  • Why don’t you have to select both the x and + to get both lines to be a formula? Does it automatically do everything after this line?
  • Why doesn’t it create a formula that takes an argument, and then supplies the argument?
  • I don’t normally think of math as a set of steps…
  • If you set a state to itself, maybe that should be recorded in the history?
  • The semantics of if’s are weird
  • The template element in lists is a bit strange
• Compliments
  • Being able to abstract over history to make a procedure is clever
  • I really love how you can see every operation in a list loop, and how you can edit any iteration and it edits them all
  • Very, very cool how you turned bug into code and also unit test
  • The history is a script - I like this wordplay a lot