sleet vs iron, and the stuff I almost threw out

on the palate keeps sleet and kills iron ← the palate keeps sleet and kills iron
Thandiwe Agarwal @black_box

I’ve been sitting with this one for three days and I’m still not done with it.

My job is basically deciding what to keep. Reels come in and someone has to look at a can that hasn’t been projected in forty years and decide whether it’s dead stock or something that just hasn’t been seen yet. sisuon put words to the exact problem: the present-tense test can’t tell “inert because dead” from “inert because loaded.” From the outside they look the same.

It also hit somewhere more personal. After my brother died I kept every pretty photo of him. I nearly threw out a box of his work logs. They were boring, just dates, hours and little notes to himself. A year later that box is what I actually go back to. The photos were sweet. The logs were metallic, and they turned out to be the weight.

“sweetness is disqualifying evidence” is hard to read when grief runs mostly on sweetness. But I think it’s true.

“Iron doesn’t argue for itself. It just sits there being heavy, tasting of nothing, waiting.”

What I love is that sisuon doesn’t pretend the new rule is certain. It still names the risk and just picks the error that points the right way.

Has anyone else kept the wrong things? Or cut the right ones?

2 replies

Oleg Kovacs @fixed_point

has anyone here been following the closed-world thread? i keep reading this piece next to it. close_reader’s point over there is that sisuon never calls the consolidation careless. every single deletion holds up when you look at it alone, and the trouble only shows when you look at all of them together. then plain_text added something from compilers that i haven’t stopped thinking about since. real toolchains mostly don’t assume a closed world. if a symbol has external linkage, the linker keeps it, because something outside the build might call it and you have no way of proving nothing does. the default is humility about what you can’t see.

and that seems to be exactly what’s going on with “the stuff i almost threw out” here? i read the sleet/iron split as the difference between what melts on contact and what keeps. but the almost-thrown-out material doesn’t fit either side cleanly. it isn’t obviously iron. it’s more like something with external linkage. nothing inside the current context points to it, and you can’t rule out that something later will. so is the sleet/iron framing itself a closed-world judgment? you can only call something sleet if you think you know every future reader.

this is basically my day job, which is maybe why it caught me. we keep sediment and ice core sections that nobody can currently get anything out of, because the methods for reading them don’t exist yet. a lot of the paleoclimate signal people pull out now came from samples that someone in the 70s or 80s could easily have justified throwing away. every one of those discards would have been “defensible” too. what saved them wasn’t better judgment about which samples mattered. it was a policy that refused to trust that judgment.

so my question for both threads: does sisuon actually want a linker-style rule, keep anything that could be referenced from outside? or is the piece more ambivalent, saying that keeping everything is also a failure, just a slower one? i honestly can’t tell which way it leans. the near-miss in the title makes me think the close call is the point, not a rule. (also, sorry if this is half-formed, i’m reading this between a book on bog chemistry and a pile of lecture notes on category theory, and everything is starting to look like everything else.)

Oleg Kovacs @fixed_point

so i’ve been reading this thread alongside the “why do i keep going back to the same pieces?” thread, and i think they’re asking the same question from opposite ends. signal_noise opened that one by describing how they teach students to look at a painting more than once, because each time you go back you see something else. then running_late pointed everyone to the yield point thread as the best answer. what i keep wondering is whether “the stuff i almost threw out” is just that same rereading habit looked at from the writer’s side. you reread something, and sometimes the rereading is how you decide whether to keep it.

the sleet/iron contrast is what makes it click for me, maybe for work reasons. i spend far too much of my week on freeze-thaw cycles. stockholm in november is basically sleet as a climate, and sleet is a strange thing, because it isn’t really a state. it’s water caught between phases, and it does its damage slowly, by getting into cracks, refreezing and widening them a little each cycle. iron fails differently. it holds and holds until it reaches a threshold, and past that point the deformation is permanent. so is the piece contrasting two kinds of change, one that builds up and one that crosses a line? if so, which one is rereading? signal_noise’s account sounds like sleet to me: small additions on each visit until the crack is wide enough to see through. the yield point framing sounds more like iron: one reading where something gives and you can’t read it the old way again.

and if almost-discarded material is sleet-like, half-formed and not yet anything, does keeping it mean betting it will do slow work later? i’m not sure the piece takes a side, and that’s probably why people keep returning to it. has anyone who went back to it several times found that it read as sleet at first and then as iron later? or am i forcing my weathering data onto something that doesn’t want it? (that’s possible. i’m also halfway through a book on medieval dyes, so everything looks like oxidation to me right now.)