change misuse of "idempotent" to something else

[michaelficarra]

There are occasions throughout the spec where the word "idempotent" is used to mean something other than its common usage in mathematics. In these cases, the spec is trying to claim that a host-defined procedure indeed behaves as a mathematical function. For example,

This operation must be idempotent if it completes normally. Each time it is called with a specific exportName, resolveSet pair as arguments it must return the same result.

Confusingly, there are also instances that are consistent with common usage.

The ToInt32 abstract operation is idempotent: if applied to a result that it produced, the second application leaves that value unchanged.

I think we can just remove the incorrect usage, since the intended meaning is always described immediately after. If someone has a better term or phrasing to use as an alternative, I'd love to hear it.

[getify]

I believe there are two "common usages" of Idempotent... one is the mathematical sense, like ABS(-1) === ABS(ABS(-1)), and the other is the programming sense, like x = 4; being equivalent to x = 4; x = 4;.

FWIW, I cover the duality of the concept of "Idempotence" in one of my book chapters.

[getify]

I believe there are two "common usages" of Idempotent... one is the mathematical sense, like ABS(-1) === ABS(ABS(-1)), and the other is the programming sense, like x = 4; being equivalent to x = 4; x = 4;.

FWIW, I cover the duality of the concept of "Idempotence" in one of my book chapters.

[tabatkins]

Yes, both definitions of the word are in common usage; for example, it's commonly stated that GET requests must be "idempotent", while PUSH requests don't have to be.

[tabatkins]

Yes, both definitions of the word are in common usage; for example, it's commonly stated that GET requests must be "idempotent", while PUSH requests don't have to be.

[ljharb]

The only definition i really ever encounter is the REST one - ie, what GET requests must be.

Can you elaborate on what’s incorrect about the current usage?

[ljharb]

The only definition i really ever encounter is the REST one - ie, what GET requests must be.

Can you elaborate on what’s incorrect about the current usage?

[michaelficarra]

@ljharb "The output of P is determined wholly by its parameters" and "if applied to any of its own outputs, it behaves as the identity function" are two very different concepts. We use the same word to mean both of those things within the same document. Something must change.

[michaelficarra]

@ljharb "The output of P is determined wholly by its parameters" and "if applied to any of its own outputs, it behaves as the identity function" are two very different concepts. We use the same word to mean both of those things within the same document. Something must change.

[ljharb]

I would prefer the former definition; the second one doesn't seem useful in a context where (unlike mathematics) equality is determined referentially instead of by value.

[ljharb]

I would prefer the former definition; the second one doesn't seem useful in a context where (unlike mathematics) equality is determined referentially instead of by value.

[michaelficarra]

@ljharb Actually, ToInt32, one of the abstract operations that uses this term according to the common mathematical usage, operates on (or at least returns) mathematical values. Mathematical values are of course compared by value.

3. Let int be the mathematical value that is the same sign as number and whose magnitude is floor(abs(number)).

[michaelficarra]

@ljharb Actually, ToInt32, one of the abstract operations that uses this term according to the common mathematical usage, operates on (or at least returns) mathematical values. Mathematical values are of course compared by value.

3. Let int be the mathematical value that is the same sign as number and whose magnitude is floor(abs(number)).

[ljharb]

@michaelficarra fair point; but in that case I think both definitions are useful.

[ljharb]

@michaelficarra fair point; but in that case I think both definitions are useful.

[robpalme]

Could we differentiate the two cases by referring to them as "deterministic" vs "idempotent"?

…

On Wed, 7 Mar 2018, 20:22 Jordan Harband, ***@***.***> wrote: @michaelficarra <https://github.com/michaelficarra> fair point; but in that case I think both definitions are useful. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1133 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGni9bBL-bVmX8Hl3EQHyoHDtya9XfS0ks5tcEFzgaJpZM4Sg348> .

[robpalme]

Could we differentiate the two cases by referring to them as "deterministic" vs "idempotent"?

…

On Wed, 7 Mar 2018, 20:22 Jordan Harband, ***@***.***> wrote: @michaelficarra <https://github.com/michaelficarra> fair point; but in that case I think both definitions are useful. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1133 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGni9bBL-bVmX8Hl3EQHyoHDtya9XfS0ks5tcEFzgaJpZM4Sg348> .

[michaelficarra]

@robpalme That would make me happy.

[michaelficarra]

@robpalme That would make me happy.

[claudepache]

Could we differentiate the two cases by referring to them as
"deterministic" vs "idempotent"?

x++ is deterministic but not idempotent (in the computer science meaning).
x → x+1 is deterministic but not idempotent (in the mathematical meaning).

[claudepache]

Could we differentiate the two cases by referring to them as
"deterministic" vs "idempotent"?

x++ is deterministic but not idempotent (in the computer science meaning).
x → x+1 is deterministic but not idempotent (in the mathematical meaning).

[michaelficarra]

@claudepache Those are expressions with effects. What would be an example of a function that you would describe as "deterministic", but which would violate what we're trying to describe here?

edit: We could be even more pedantic and say "deterministic with respect to its parameters".

[michaelficarra]

@claudepache Those are expressions with effects. What would be an example of a function that you would describe as "deterministic", but which would violate what we're trying to describe here?

edit: We could be even more pedantic and say "deterministic with respect to its parameters".

[rossberg]

@getify:

I believe there are two "common usages" of Idempotent... one is the mathematical sense, like ABS(-1) === ABS(ABS(-1)), and the other is the programming sense, like x = 4; being equivalent to x = 4; x = 4;

The second is also the mathematical sense. The term is used there because semantically, you can think about the state of a program as an abstract value that is "computed" by executing statements. The semantics of a statement can be seen as a function from state to state. And for a statement like the above, this function is idempotent.

More concretely, the semantics of "x = 4", often written [[x = 4]], is a function where the result is the same state value as the arg but with its internal mapping for x changed. And if you have an initial state S, then that resulting state S' = [[x=4]] (S) is always the same as the one resulting from [[x=4]] ([[x=4]] (S)).

[rossberg]

@getify:

I believe there are two "common usages" of Idempotent... one is the mathematical sense, like ABS(-1) === ABS(ABS(-1)), and the other is the programming sense, like x = 4; being equivalent to x = 4; x = 4;

The second is also the mathematical sense. The term is used there because semantically, you can think about the state of a program as an abstract value that is "computed" by executing statements. The semantics of a statement can be seen as a function from state to state. And for a statement like the above, this function is idempotent.

More concretely, the semantics of "x = 4", often written [[x = 4]], is a function where the result is the same state value as the arg but with its internal mapping for x changed. And if you have an initial state S, then that resulting state S' = [[x=4]] (S) is always the same as the one resulting from [[x=4]] ([[x=4]] (S)).

[rossberg]

Markup messed up the notation. I meant to write that the state S' = [[x=4]] (S) is always the same as the one resulting from [[x=4]] ([[x=4]] (S)).

[rossberg]

Markup messed up the notation. I meant to write that the state S' = [[x=4]] (S) is always the same as the one resulting from [[x=4]] ([[x=4]] (S)).