Talk 5: Cameron Swords

[rrnewton]

No description provided.

[jasonhemann]

You'll get this from everyone, but on grey background red parens are invisible. Was that on purpose to de-emphasize the parens?

[ccshan]

Visible parentheses please.

[rrnewton]

Another word choice one -- I might vote for "seminal" over iconic.

[samth]

• Myriad can also be an adjective
• I like the parens.

[samth]

Include the full author list (for example, Findler & Felleisen).

[rrnewton]

I like the rubric of monitoring strategies -- are these your names (semi, async, temporal...) or someone else's?

[rrnewton]

Did you say Haskell was "semi eager" monitoring? I was having trouble understanding which of these compose with the others, if any.

[rrnewton]

Re: Parens -- wow, what a showcase in how horrible projectors are. They looked great on the TV.

[ccshan]

Word choice? Avoid "slave".

[jasonhemann]

I might have missed it. Was "best effort" explained?

[rrnewton]

Future contracts sound cool -- do those actually work / work well in Racket?

"If it gets an error, it brings the whole threadpool to a stop" -- really? Shouldn't an error inside a future not do anything until the future is forced?

[samth]

• Future/semi motioning slide: show the 2 + [] context.

[pnwamk]

• if we're de-emphasizing Lisp-isms in the code, go ahead and get rid of the commas before the numbers in the quoted sexps
• maybe put 'sync' or 'barrier' or 'wait' or something on the horizontal lines on your diagrams?
• v_c forgot the _c on slide right before semi-eager monitoring

[samth]

• Need to spend 3x as much time on the code for the different strategies.

[jasonhemann]

One of the parens in your oval-of-strategies stood out to me. Talk about picayune complaints, but ))) the 1st and 2nd were closer than the 3rd.

[pnwamk]

(actually several of the 'v' contract values should be 'v_c' in the last set of slides)

[rrnewton]

raise & catch -- PLUS spawn. How do raise and catch interact with spawn? Spawn is for future creation, right? In that case exceptions are lazy, observed at the point of the force.

On the slides after that though, I saw delay but not spawn... where did spawn go?

[rrnewton]

"maintains the tree's aspmytotic guarantees"?

Minor nitpick -- the things with asymptotic guarantees are functions over the trees (the ADT methods) not the data itself, right?

[pnwamk]

I don't understand why the last slide has that (albeit very pretty looking) table of the 4 strategies and their differences. It feels a little 'heavy' for a takeaway snapshot IMHO.

[rrnewton]

Dan: "should you be using monads"?

I.e. you're passing the strategies all over the place... should that threaded through by a monad?

Cameron: the point is to have explicit control.

(This sounds like an argument for default/optional arguments or whatever.)

[rrnewton]

Ryan Scott: one thing that was incredible to me is that you have these different eval strategies but you can use them in tandem.

[sabry]

bst — property you’re checking is NOT balanced

say something about properties: faithfulness, blah

pointer to point at screen

get rid of two slides for ‘building the separation’ just present it with syntax

that’s it! Nice slide (Strategy Strategy)

Ouch. Had to slip important slide

better explanation of tree fullness… can say less detail but more high level intuition about collection of contracts communicating together to establish global invariants

when asked questions: if they are way out there, don’t speculate too much. ok to say “will talk later…”

composing strategies. good point. dan’s point I think is that you can hide the management of strategies in a monadic interface

your answers are too rambling though… give them to the point, concise

you never said you have an implementation. Say it and highlight

composable strategies: i haven’t seen two that don’t compose. not a good answer.

[vikraman]

I think the bst contract had a typo in the last line.

[cgswords]

Thank you everyone for your feedback!