search

junyechen1996 / draft-chen-cfrg-vdaf-pine

VDAF to support aggregating real number vectors with L2-norm bound
Other
4 stars 0 forks source link

Clean up some texts for wraparound checks #72

Closed junyechen1996 closed 6 months ago

junyechen1996 commented 6 months ago

We have conflicting references of "wraparound check results", e.g., "Encoding range-checked wraparound check results", which are the bits of the range-checked result in wraparound check, and we also have "wraparound check results", that we typically mean the dot products from running wraparound tests.

How about we use "check" on the entire wraparound "check" section, i.e., it is checking for wraparound, similar to L2-norm "check". And we also have the quadratic "check" and success count "check" in wraparound checks. For the dot product computation, we will refer to it as wraparound "test". I think we already did this in a couple of places, but just weren't consistent about it.

junyechen1996 commented 6 months ago

@cjpatton If you like this idea, I can make a pass over the variables (mainly num_wr_checks -> num_wr_tests) and reference implementations.

cjpatton commented 6 months ago

We have conflicting references of "wraparound check results", e.g., "Encoding range-checked wraparound check results", which are the bits of the range-checked result in wraparound check, and we also have "wraparound check results", that we typically mean the dot products from running wraparound tests.

How about we use "check" on the entire wraparound "check" section, i.e., it is checking for wraparound, similar to L2-norm "check". And we also have the quadratic "check" and success count "check" in wraparound checks. For the dot product computation, we will refer to it as wraparound "test". I think we already did this in a couple of places, but just weren't consistent about it.

I'd prefer to use one term: either "check" or "test". Mixing them would be confusing.

Instead, I think our goal should be to make sure the distinction is clear in context: one is encoded in the in the input (shares), the other is computed from the input (shares).

junyechen1996 commented 6 months ago

Instead, I think our goal should be to make sure the distinction is clear in context: one is encoded in the in the input (shares), the other is computed from the input (shares).

Got it. I reworked this PR and tried to clarify the distinction between them.

junyechen1996 commented 6 months ago

Squashed.