incommensurability
Nov. 20th, 2001 01:02 am![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
eubI had a three-hour conversation with a visiting prof -- trying to explain what I do, and then we switched roles. It was mind-bendingly difficult.
I explain the three type constructors I use, and what structure they introduce. She asks what their semantics are. I say they don't really have semantics, it depends on how you want to interpret them. She kind of insists that they do. I don't see how.
She explains how her system is based on events with "process" bodies that are, I think, lambda expressions in terms of logical start time. I try to ask about access to free variables with a scenario of a global control envelope, foo, that instantiated notes sample current values from. The question I'm asking, "how does the process access foo's current value?", does not appear to make any sense. Finally I think we've got it: we agree that foo has to be a process that takes a time and returns a scalar value. I say, okay, then the note's process is (\t. blah blah (foo t) blah)? No, you don't do this stuff in the process, you magic it into an attribute that hangs on the side of the note.
This all was probably so hard because she works in a tangentially-related area, so we talk about some of the same things, but we bring our assumptions to them from different places. Someone from an alternate universe is good to talk to. I was never certain if what I discovered was the same as what she was saying, but oh well.
For one thing, I could make my type system suitable ground for a semantics, if I refined it further, so as to differentiate the cases where it would otherwise be ambiguous. I thought about that at the very beginning, but buried it because it seemed too hard.
The two of us have very different ideas about how much musical know-how to try to build in to the system. She's working towards a grand unified theory of musical structure (from the note level upwards -- e.g. no timbre). I deliberately give the programmer mindless structured data and absolute freedom to screw with it.
My feeling is that it's better to start with something I know is usable and refine the type system later. If I built much more in, I'd be scared of hardwiring in my compositional habits and assumptions without even seeing them. There are to many high-level music-description systems that only their writer can wrap their head around.
I explain the three type constructors I use, and what structure they introduce. She asks what their semantics are. I say they don't really have semantics, it depends on how you want to interpret them. She kind of insists that they do. I don't see how.
She explains how her system is based on events with "process" bodies that are, I think, lambda expressions in terms of logical start time. I try to ask about access to free variables with a scenario of a global control envelope, foo, that instantiated notes sample current values from. The question I'm asking, "how does the process access foo's current value?", does not appear to make any sense. Finally I think we've got it: we agree that foo has to be a process that takes a time and returns a scalar value. I say, okay, then the note's process is (\t. blah blah (foo t) blah)? No, you don't do this stuff in the process, you magic it into an attribute that hangs on the side of the note.
This all was probably so hard because she works in a tangentially-related area, so we talk about some of the same things, but we bring our assumptions to them from different places. Someone from an alternate universe is good to talk to. I was never certain if what I discovered was the same as what she was saying, but oh well.
For one thing, I could make my type system suitable ground for a semantics, if I refined it further, so as to differentiate the cases where it would otherwise be ambiguous. I thought about that at the very beginning, but buried it because it seemed too hard.
The two of us have very different ideas about how much musical know-how to try to build in to the system. She's working towards a grand unified theory of musical structure (from the note level upwards -- e.g. no timbre). I deliberately give the programmer mindless structured data and absolute freedom to screw with it.
My feeling is that it's better to start with something I know is usable and refine the type system later. If I built much more in, I'd be scared of hardwiring in my compositional habits and assumptions without even seeing them. There are to many high-level music-description systems that only their writer can wrap their head around.
