From 0e5de7766173a2812183c5c26759c1c27e8d8e40 Mon Sep 17 00:00:00 2001
From: Arun Isaac
Date: Wed, 2 Oct 2024 01:07:28 +0100
Subject: propnet: Explain in detail why we never cancel running propagators.

* ravanan/propnet.scm (poll-propnet): Add detailed comment elaborating
on why we never cancel or forget about running propagators.
---
 ravanan/propnet.scm | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/ravanan/propnet.scm b/ravanan/propnet.scm
index ca5c629..e6155d8 100644
--- a/ravanan/propnet.scm
+++ b/ravanan/propnet.scm
@@ -210,8 +210,15 @@ add to the inbox."
                 cell-values-inbox
                 other-propagators
                 ;; We don't need to cancel or forget about previous runs of the
-                ;; same propagator because cells only "accumulate" information;
-                ;; they never remove it.
+                ;; same propagator because cells only *accumulate* information;
+                ;; they never remove it. Any previous runs of the same
+                ;; propagator will only *add to* the information in the output
+                ;; cells. Previous runs may be closer to completion and taking
+                ;; advantage of their output may allow later stages to start
+                ;; running sooner, thus improving throughput. In our CWL
+                ;; application of propnets, this will never result in the same
+                ;; step being recomputed; so this approach does not come at a
+                ;; higher computational cost.
                 (append (maybe-alist
                          (cons (propagator-name propagator)
                                (activate-propagator
-- 
cgit v1.2.3