diff --git a/gzip_accept_encoding_htab.t b/gzip_accept_encoding_htab.t
new file mode 100644
index 00000000..76b09f90
--- /dev/null
+++ b/gzip_accept_encoding_htab.t
@@ -0,0 +1,125 @@
+#!/usr/bin/perl
+
+# (C) Maxim Dounin
+
+# Tests for nginx gzip filter module HTAB handling in Accept-Encoding.
+#
+# Verifies that a horizontal tab (HTAB, 0x09) is treated as whitespace
+# between comma-separated tokens in the Accept-Encoding request header,
+# matching the SP behavior already supported by ngx_http_gzip_accept_encoding().
+#
+# Per RFC 9110 §5.6.3, the Accept-Encoding field value is a "#rule" list of
+# codings; the surrounding field grammar (RFC 9110 §5.6.2) defines optional
+# whitespace as SP / HTAB. The companion HTTP/1 parser already accepts HTAB
+# alongside SP at the field-value boundary (RFC 9112 §2.2), and Vadim's
+# nginx#1577 taught the multi-header parser to accept HTAB as well. The
+# gzip_accept_encoding helper in ngx_http_core_module.c carries its own
+# boundary check, so it must be updated to match.
+#
+# The gap is the single-character path that runs after the gzip token has
+# been matched: the switches in the function only carry case ' ' (space) and
+# exit on the bare comma that ends the list, so HTAB falls into the default
+# branch and the encoding is rejected. Test 1 below covers the case where
+# the comma is hit before HTAB, which therefore works correctly today and
+# is held as a regression sentinel.
+#
+# Code PR: nginx/nginx#<companion>
+
+###############################################################################
+
+use warnings;
+use strict;
+
+use Test::More;
+
+BEGIN { use FindBin; chdir($FindBin::Bin); }
+
+use lib 'lib';
+use Test::Nginx qw/ :DEFAULT :gzip /;
+
+###############################################################################
+
+select STDERR; $| = 1;
+select STDOUT; $| = 1;
+
+my $t = Test::Nginx->new()->has(qw/http gzip/)->plan(7);
+
+$t->write_file_expand('nginx.conf', <<'EOF');
+
+%%TEST_GLOBALS%%
+
+daemon off;
+
+events {
+}
+
+http {
+    %%TEST_GLOBALS_HTTP%%
+
+    server {
+        listen       127.0.0.1:8080;
+        server_name  localhost;
+
+        location / {
+            gzip on;
+        }
+    }
+}
+
+EOF
+
+$t->write_file('index.html', 'X' x 64);
+
+$t->run();
+
+###############################################################################
+
+# Baseline: SP / no whitespace between tokens is recognised.
+like(http_ae('gzip, deflate'), qr/Content-Encoding: gzip/,
+    'gzip enabled with SP between tokens');
+
+like(http_ae('gzip,deflate'), qr/Content-Encoding: gzip/,
+    'gzip enabled with no whitespace between tokens');
+
+# Regression sentinel: HTAB after the comma is recognised because the comma
+# ends the token scan before HTAB is inspected.
+like(http_ae("gzip,\tdeflate"), qr/Content-Encoding: gzip/,
+    'gzip enabled with HTAB after comma');
+
+# RFC 9110 §5.6.2 requires comma separators between elements; whitespace
+# alone is not enough. The patch accepts HTAB as whitespace but does not
+# change the rejection of malformed input that lacks a comma.
+unlike(http_ae("gzip\tdeflate"), qr/Content-Encoding: gzip/,
+    'malformed Accept-Encoding without comma is rejected');
+
+TODO: {
+local $TODO = 'gzip accept_encoding does not recognise HTAB';
+
+# HTAB before the comma: Accept-Encoding: gzip\t,deflate.
+like(http_ae("gzip\t,deflate"), qr/Content-Encoding: gzip/,
+    'gzip enabled with HTAB before comma');
+
+# HTAB inside the quantity: gzip;\tq=0.5.
+like(http_ae("gzip;\tq=0.5"), qr/Content-Encoding: gzip/,
+    'gzip enabled with HTAB before q-value');
+
+# HTAB after the q-value, before the comma: gzip;q=0.5\t,deflate.
+like(http_ae("gzip;q=0.5\t,deflate"), qr/Content-Encoding: gzip/,
+    'gzip enabled with HTAB after q-value before comma');
+
+}
+
+###############################################################################
+
+sub http_ae {
+	my ($ae) = @_;
+	return http(<<EOF);
+GET / HTTP/1.1
+Host: localhost
+Connection: close
+Accept-Encoding: $ae
+
+EOF
+}
+
+###############################################################################