search

ietf-wg-ppm / draft-ietf-ppm-dap-taskprov

Define DAP extension for in-band task provisioning.
Other
3 stars 4 forks source link

Change dp_config to opaque #58

Closed junyechen1996 closed 10 months ago

junyechen1996 commented 10 months ago

Also remove length hints in TLS variants that are mandatory to configure a task.

junyechen1996 commented 10 months ago

I'm mainly trying to align with our implementations, but I'd welcome feedback on the texts here.

cjpatton commented 10 months ago

I think this is a regression. The goal of encoding the length of query type, vdaf type, and dp type is so that we can decode a variant we don't recognize. The goal was to separate parsing from protocol logic that is implementation specific.

I think this is a regression.

Thanks @junyechen1996 for pointing this out.

cjpatton commented 10 months ago

@wangshan you might remember we discussed this a few months ago.

wangshan commented 10 months ago

@cjpatton this isn't a regression, what we had before was a DpConfig with no length so decoder may not know how to skip it. What we write here uses opaque dp_config<1..2^16-1> which encodes the length in the variant length array already. When decoding, you either know how to decode dp_config or know how long it is by reading the first 2 bytes.

cjpatton commented 10 months ago

@cjpatton this isn't a regression, what we had before was a DpConfig with no length so decoder may not know how to skip it. What we write here uses opaque dp_config<1..2^16-1> which encodes the length in the variant length array already. When decoding, you either know how to decode dp_config or know how long it is by reading the first 2 bytes.

The regression is: There is no length prefix for VdafConfig or QueryConfig. DpConfig is fine.