[CFRG] Should we spell out the star topology for VDAFs?

[Christopher Patton <[e-mail]>]

Hi PPM+CFRG,

I'd like to get some feedback on what to do about this issue for the VDAF
draft:
https://github.com/cfrg/draft-irtf-cfrg-vdaf/issues/261

Prior to DAP draft 05, we envisioned a "star topology" in which there is a
single leader and any number of helpers. The idea was that for each round
of preparation, the leader would broadcast the prep message to each of the
helpers: each helper would respond with its prep share for the next round,
which the leader would use to compute the prep message for the next round,
and so on. In draft 05 we decided to specialize the protocol for a single
leader and helper and adopt the ping-pong topology [1], which requires
fewer round trips over the network to finish.

In theory, a future specification could execute a VDAF in a star topology,
so it may be useful to spell this out. Note however that not all VDAFs
support more than two aggregators: Prio3 does, but Poplar1 doesn't.

The VDAF draft spells out ping ponging in full detail. This is necessary
because the DAP draft consumes this API directly. Do folks think we should
spell out the star topology in the same level of detail?

Since we don't currently have a use case for it, I don't think it's
necessary to spell out in full detail. I suggest we just sketch it instead.

Thanks!
Chris P.

[1] https://www.ietf.org/archive/id/draft-irtf-cfrg-vdaf-10.html#section-5.8