diff --git a/Makefile b/Makefile
index 62583726a..90088c08c 100644
--- a/Makefile
+++ b/Makefile
@@ -271,9 +271,9 @@ ifeq ($(STATIC),1)
 # For MacOS, Jacob Rapoport <jacob@rumblemonkey.com> changed this to:
 #  -L/usr/local/lib -lsqlite3 -lz -Wl,-lm -lpthread -ldl $(COVFLAGS)
 # But that doesn't static link.
-LDLIBS = -L$(CPATH) -Wl,-dn $(SQLITE3_LDLIBS) -lz -Wl,-dy -lm -lpthread -ldl $(COVFLAGS)
+LDLIBS = -L$(CPATH) -Wl,-dn $(SQLITE3_LDLIBS) -Wl,-dy -lm -lpthread -ldl $(COVFLAGS)
 else
-LDLIBS = -L$(CPATH) -lm $(SQLITE3_LDLIBS) -lz $(COVFLAGS)
+LDLIBS = -L$(CPATH) -lm $(SQLITE3_LDLIBS) $(COVFLAGS)
 endif
 
 # If we have the postgres client library we need to link against it as well
diff --git a/common/decode_array.c b/common/decode_array.c
index 86119d4e5..0f9a8310b 100644
--- a/common/decode_array.c
+++ b/common/decode_array.c
@@ -2,25 +2,6 @@
 #include <ccan/cast/cast.h>
 #include <common/decode_array.h>
 #include <wire/peer_wire.h>
-#include <zlib.h>
-
-static u8 *unzlib(const tal_t *ctx, const u8 *encoded, size_t len)
-{
-	/* http://www.zlib.net/zlib_tech.html gives 1032:1 as worst-case,
-	 * which is 67632120 bytes for us.  But they're not encoding zeroes,
-	 * and each scid must be unique.  So 1MB is far more reasonable. */
-	unsigned long unclen = 1024*1024;
-	int zerr;
-	u8 *unc = tal_arr(ctx, u8, unclen);
-
-	zerr = uncompress(unc, &unclen, encoded, len);
-	if (zerr != Z_OK)
-		return tal_free(unc);
-
-	/* Truncate and return. */
-	tal_resize(&unc, unclen);
-	return unc;
-}
 
 struct short_channel_id *decode_short_ids(const tal_t *ctx, const u8 *encoded)
 {
@@ -43,11 +24,7 @@ struct short_channel_id *decode_short_ids(const tal_t *ctx, const u8 *encoded)
 	type = fromwire_u8(&encoded, &max);
 	switch (type) {
 	case ARR_ZLIB_DEPRECATED:
-		encoded = unzlib(tmpctx, encoded, max);
-		if (!encoded)
-			return NULL;
-		max = tal_count(encoded);
-		/* fall thru */
+		return NULL;
 	case ARR_UNCOMPRESSED:
 		scids = tal_arr(ctx, struct short_channel_id, 0);
 		while (max) {
@@ -86,11 +63,7 @@ bigsize_t *decode_scid_query_flags(const tal_t *ctx,
 	 */
 	switch (qf->encoding_type) {
 	case ARR_ZLIB_DEPRECATED:
-		encoded = unzlib(tmpctx, encoded, max);
-		if (!encoded)
-			return NULL;
-		max = tal_count(encoded);
-		/* fall thru */
+		return NULL;
 	case ARR_UNCOMPRESSED:
 		flags = tal_arr(ctx, bigsize_t, 0);
 		while (max)
@@ -120,11 +93,7 @@ decode_channel_update_timestamps(const tal_t *ctx,
 	 * query_short_channel_ids_tlvs! */
 	switch (timestamps_tlv->encoding_type) {
 	case ARR_ZLIB_DEPRECATED:
-		encoded = unzlib(tmpctx, encoded, max);
-		if (!encoded)
-			return NULL;
-		max = tal_count(encoded);
-		/* fall thru */
+		return NULL;
 	case ARR_UNCOMPRESSED:
 		ts = tal_arr(ctx, struct channel_update_timestamps, 0);
 		while (max) {
diff --git a/devtools/mkencoded.c b/devtools/mkencoded.c
index 368b66993..6ff556be9 100644
--- a/devtools/mkencoded.c
+++ b/devtools/mkencoded.c
@@ -5,7 +5,6 @@
 #include <ccan/str/hex/hex.h>
 #include <common/decode_array.h>
 #include <stdio.h>
-#include <zlib.h>
 
 static NORETURN void usage(void)
 {
@@ -43,20 +42,7 @@ int main(int argc, char *argv[])
 	if (encoding == ARR_UNCOMPRESSED)
 		printf("%02x%s\n", encoding, tal_hex(NULL, data));
 	else if (encoding == ARR_ZLIB_DEPRECATED) {
-		/* https://www.zlib.net/zlib_tech.html:
-		 *   the only expansion is an overhead of five bytes per 16 KB
-		 *   block (about 0.03%), plus a one-time overhead of six bytes
-		 *   for the entire stream.
-		 */
-		unsigned long compressed_len = tal_bytelen(data)
-			+ 6 + 5*(tal_bytelen(data) + 16384)/16384;
-		u8 *z = tal_arr(NULL, u8, compressed_len);
-		if (compress2(z, &compressed_len, data, tal_bytelen(data),
-			      Z_DEFAULT_COMPRESSION) != Z_OK)
-			errx(1, "Error compressing stream?");
-
-		printf("%02x%s\n", encoding,
-		       tal_hexstr(NULL, z, compressed_len));
+		errx(1, "ZLIB compression deprecated");
 	} else {
 		errx(1, "Unknown encoding %u", encoding);
 	}
diff --git a/gossipd/queries.c b/gossipd/queries.c
index a282abddf..200a6f82e 100644
--- a/gossipd/queries.c
+++ b/gossipd/queries.c
@@ -14,7 +14,6 @@
 #include <gossipd/gossipd_wiregen.h>
 #include <gossipd/gossmap_manage.h>
 #include <gossipd/queries.h>
-#include <zlib.h>
 
 static u32 dev_max_encoding_bytes = -1U;
 
diff --git a/tools/headerversions.c b/tools/headerversions.c
index 00435e867..d2a0f85db 100644
--- a/tools/headerversions.c
+++ b/tools/headerversions.c
@@ -14,15 +14,12 @@
 # define IF_SQLITE3(...)
 #endif
 #include <unistd.h>
-#include <zlib.h>
 
 static const char template[] =
 	"/* Generated file by tools/headerversions, do not edit! */\n"
 	IF_SQLITE3("/* SQLITE3 version: %u */\n")
-	"/* ZLIB version: %s */\n"
 	"#include <ccan/err/err.h>\n"
 	IF_SQLITE3("#include <sqlite3.h>\n")
-	"#include <zlib.h>\n"
 	"\n"
 	"static inline void check_linked_library_versions(void)\n"
 	"{\n"
@@ -36,10 +33,6 @@ static const char template[] =
 	"		errx(1, \"SQLITE major version mismatch: compiled %%u, now %%u\",\n"
 	"		     SQLITE_VERSION_NUMBER, sqlite3_libversion_number());\n"
 	)
-	"	/* zlib documents that first char alters ABI. Kudos! */\n"
-	"	if (zlibVersion()[0] != ZLIB_VERSION[0])\n"
-	"		errx(1, \"zlib version mismatch: compiled %%s, now %%s\",\n"
-	"		     ZLIB_VERSION, zlibVersion());\n"
 	"}\n";
 
 int main(int argc, char *argv[])
@@ -57,8 +50,7 @@ int main(int argc, char *argv[])
 		err(1, "Reading %s", argv[1]);
 
 	new = tal_fmt(NULL, template,
-		      IF_SQLITE3(sqlite3_libversion_number(),)
-		      zlibVersion());
+		      IF_SQLITE3(sqlite3_libversion_number()));
 	if (!file || !streq(new, file)) {
 		int fd = open(argv[1], O_TRUNC|O_WRONLY|O_CREAT, 0666);
 		if (fd < 0)