#general
- e
enough-analyst-54434
10/07/2015, 11:32 PMThis was well after that refactor - b
bored-art-40741
10/07/2015, 11:33 PMRight, hence slipping - b
bored-art-40741
10/07/2015, 11:33 PMI was just mentioning I agree that we should have type barriers in place here - b
bored-art-40741
10/07/2015, 11:33 PMBut it's hard - e
enough-analyst-54434
10/07/2015, 11:33 PMWell, I'm not sure about slip, in all languages I've used, reifing types at the boundaries has always emerged as a principle - e
enough-analyst-54434
10/07/2015, 11:34 PMtake input, immediately firm it up into the internal model - e
enough-analyst-54434
10/07/2015, 11:34 PMthe more you delay, the more you get screwed by yourself - w
witty-crayon-22786
10/07/2015, 11:34 PMyea. - b
bored-art-40741
10/07/2015, 11:34 PMAgreed, and that was in fact done for specs - b
bored-art-40741
10/07/2015, 11:34 PMAnd then there came other ways to pass in specs, and they weren't normalized immediately on input - b
bored-art-40741
10/07/2015, 11:35 PMThough we never actually plugged the options hole here - e
enough-analyst-54434
10/07/2015, 11:35 PMSo this has been useful - one thing we don't have and that we - meaning Bill and I - tried to establish much too late is principles for the project - w
witty-crayon-22786
10/07/2015, 11:35 PMlong story short, we agree we want all task+target inputs to be typed - e
enough-analyst-54434
10/07/2015, 11:35 PMI'll file an issue and by close of Novemeber throw up a straw man for this for our docs - e
enough-analyst-54434
10/07/2015, 11:36 PMwe want everything to be types - w
witty-crayon-22786
10/07/2015, 11:36 PMbut that perhaps the PayloadField types should move out of Payload (to
or something)?Field - e
enough-analyst-54434
10/07/2015, 11:36 PMtyped - e
enough-analyst-54434
10/07/2015, 11:36 PMI think Payload goes away, but too early to say - w
witty-crayon-22786
10/07/2015, 11:37 PMwithout the fields, a payload is just a dict. - b
bored-art-40741
10/07/2015, 11:37 PMPayload and target get more indistinguishable in the new engine - e
enough-analyst-54434
10/07/2015, 11:37 PMyes - w
witty-crayon-22786
10/07/2015, 11:37 PMsure. but as we said, we need typed fields - e
enough-analyst-54434
10/07/2015, 11:37 PMPayload -> struct - w
witty-crayon-22786
10/07/2015, 11:37 PMand that's the actual meat of payload - e
enough-analyst-54434
10/07/2015, 11:38 PMBut, it will probably not live in targets like today is my gist - b
bored-art-40741
10/07/2015, 11:38 PMPayload is also immutable and stably fingerprintable - w
witty-crayon-22786
10/07/2015, 11:38 PMyep, perfect. - b
bored-art-40741
10/07/2015, 11:38 PMAnd serializable - e
enough-analyst-54434
10/07/2015, 11:38 PMeach Planner has its own notion of what the payload is - w
witty-crayon-22786
10/07/2015, 11:38 PM@bored-art-40741: well, or naturally shippable. files probably need first class support.