From 933dbf7bc5d5532a3f2a8e18ba4afb02059aa5c1 Mon Sep 17 00:00:00 2001
From: Aanish Bhirud <47579874+baanish@users.noreply.github.com>
Date: Fri, 20 Mar 2026 03:02:25 -0400
Subject: [PATCH] fix(arx): prefer chat-safe transport lengths

---
 src/lib/payload/fragment.ts | 14 ++++++++++----
 tests/arx-codec.test.ts     | 22 ++++++++++++++++++++++
 2 files changed, 32 insertions(+), 4 deletions(-)

diff --git a/src/lib/payload/fragment.ts b/src/lib/payload/fragment.ts
index 2b6edfb..516cb18 100644
--- a/src/lib/payload/fragment.ts
+++ b/src/lib/payload/fragment.ts
@@ -36,17 +36,23 @@ type CandidateFragment = {
   transportLength: number;
 };
 
+const CHAT_SAFE_ASCII_FRAGMENT_CHARS = /^[A-Za-z0-9\-._~=#]+$/;
+
 /**
- * Computes the serialized length of a fragment value as it would appear in a
- * URL after browser percent-encoding of non-ASCII characters.
- * Each non-ASCII UTF-8 byte is encoded as %XX (3 chars per byte).
+ * Computes the serialized length of a fragment value after conservative transport escaping.
+ *
+ * We count non-ASCII code points by their UTF-8 percent-encoded size, and we also treat
+ * ASCII punctuation outside the URL-unreserved fragment subset as escape-prone because many
+ * chat/link surfaces rewrite those characters even when a browser would accept them in-place.
+ * This keeps auto-selection aligned with the product's chat-safe fragment goal, allowing the
+ * `B.` base64url ARX wire shape to win when punctuation-heavy base76 would grow after sharing.
  */
 function computeTransportLength(value: string): number {
   let len = 0;
   for (let i = 0; i < value.length; i++) {
     const cp = value.codePointAt(i)!;
     if (cp < 128) {
-      len += 1;
+      len += CHAT_SAFE_ASCII_FRAGMENT_CHARS.test(value[i]) ? 1 : 3;
     } else if (cp < 0x800) {
       len += 6; // 2 UTF-8 bytes → %XX%XX
     } else if (cp < 0x10000) {
diff --git a/tests/arx-codec.test.ts b/tests/arx-codec.test.ts
index 8243ca7..eee8888 100644
--- a/tests/arx-codec.test.ts
+++ b/tests/arx-codec.test.ts
@@ -296,6 +296,28 @@ describe("arx fragment round-trip", () => {
     expect(autoHash).toContain(`v1.arx.${getActiveDictVersion()}.`);
   });
 
+  it("async arx selection can choose the chat-safe base64url wire form", async () => {
+    const bigEnvelope: PayloadEnvelope = {
+      ...envelope,
+      artifacts: [
+        {
+          id: "doc",
+          kind: "markdown",
+          filename: "doc.md",
+          content: [
+            "# Chat-safe ARX",
+            "",
+            ...Array.from({ length: 120 }, (_, index) => `- item ${index}: The quick brown fox jumps over the lazy dog.`),
+          ].join("\n"),
+        },
+      ],
+    };
+
+    const autoHash = await encodeEnvelopeAsync(bigEnvelope, { codec: "arx" });
+    expect(autoHash).toContain(`v1.arx.${getActiveDictVersion()}.B.`);
+  });
+
+
   it("decodes arx fragments when unicode payload chars are percent-escaped", async () => {
     const hash = `#${await encodeEnvelopeAsync(envelope, { codec: "arx" })}`;
     const escapedHash = hash.replace(/[^\x00-\x7F]/g, (char) => encodeURIComponent(char));