Function bao_encoded_size 

pub fn bao_encoded_size(
    total_bytes: u64,
    chunk_ranges: &RangeSet<[ChunkNum; 2]>,
) -> u64

Exact byte length of the header-less bao interleaved encoding a serving node emits for chunk_ranges of a total_bytes-byte blob — the sum of every proof node (64 bytes each) and data leaf (≤ chunk-group bytes) in response (pre-order) order (ADR 038 §Wire format).

This is the byte count that travels on cdn/client/v1: the wire carries the response-format stream (no 8-byte size header — total_bytes is already the signed StreamResponse field), so this excludes the header. It is computed from the same BaoTree + ChunkRanges the serve-side encoder walks, so it equals the emitted length byte-for-byte. The receiver uses it as the voucher-cadence / overrun / short-delivery bound, because paid bytes are wire bytes — content plus proof (ADR 038 §Payment metering) — not content bytes. chunk_ranges should be the AlignedRange::chunk_ranges both sides derive from align_range, keeping encoder and decoder in lock-step.

Cost: O(log groups) for the group-aligned contiguous range align_range produces (a closed-form count over the tree’s shape, see aligned_wire_size); every other range shape falls back to walking the encoder’s node set, which is O(groups in range). The closed form is what keeps a receiver’s wire bound cheap against a peer-claimed or ceiling-sized total_bytes in the terabytes, where a walk would visit tens of millions of groups before the first byte arrives.