v0.347.239

Settle on near and far for the two sides, and say why

The first draft mixed all three vocabularies: the voice spans said Host and
Far, the scope said host and far end, the files were host.opus and far.opus.
Near and far now runs through the file names, the voice spans and the prose,
with a section recording the reasoning so host and guest do not creep back
in.

Near-end and far-end are the standard vocabulary of echo cancellation and
VoIP, which is exactly this topology -- the near end is the microphone, the
far end is what arrives from the call and comes out of the speaker. For an
audio tool that is the right audience to be legible to.

The other two are already spoken for. local means local audio capture as
opposed to the TCP transport throughout server.zig and main.zig, in
comments, log lines and the usage text, so local/remote would give one word
two meanings in one file. And host/guest belongs to the hardware: a Vocaster
exposes a Host Microphone and a Guest Microphone for two people in the same
room, so if capsper ever captures that second mic, host/guest for the two
ends of a call leaves a host, a guest and a remote guest. Near/far is also
the only pair that survives that growth, since near covers everyone in the
room and far everyone on the call.

One weakness is accepted rather than solved and written down as such: and are terse for a human reading the transcript against the
audio. They are the honest defaults, because we do not know who the far end
is, and substituting real names is a later concern.

The one surviving remote host is the networking sense in the colon
rationale, and the naming section says so, since renaming that would be
worse than the apparent inconsistency.

Co-Authored-By: Claude Opus 5 (1M context) noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_0125wskUKzipBM5sSU5B9AZz