diff --git a/src/bitcomp.c b/src/bitcomp.c
new file mode 100644
index 0000000..bf35927
--- /dev/null
+++ b/src/bitcomp.c
@@ -0,0 +1,480 @@
+/* bit compression routines */
+
+/* (C) 2016 sysmocom s.f.m.c. GmbH by Max Suraev <msuraev@sysmocom.de>
+ *
+ * All Rights Reserved
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ */
+
+/*! \defgroup bitcomp Bit compression
+ *  @{
+ */
+
+/*! \file bitcomp.c
+ *  \brief Osmocom bit compression routines
+ */
+
+#include <stdint.h>
+#include <stdbool.h>
+#include <errno.h>
+#include <string.h>
+
+#include <osmocom/core/bitvec.h>
+#include <osmocom/core/bitcomp.h>
+
+/*
+ * Terminating codes for uninterrupted sequences of 0 and 1 up to 64 bit length
+ * according to TS 44.060 9.1.10
+ */
+static const unsigned t4_term[2][64] = {
+	{
+		0b0000110111,
+		0b10,
+		0b11,
+		0b010,
+		0b011,
+		0b0011,
+		0b0010,
+		0b00011,
+		0b000101,
+		0b000100,
+		0b0000100,
+		0b0000101,
+		0b0000111,
+		0b00000100,
+		0b00000111,
+		0b000011000,
+		0b0000010111,
+		0b0000011000,
+		0b0000001000,
+		0b00001100111,
+		0b00001101000,
+		0b00001101100,
+		0b00000110111,
+		0b00000101000,
+		0b00000010111,
+		0b00000011000,
+		0b000011001010,
+		0b000011001011,
+		0b000011001100,
+		0b000011001101,
+		0b000001101000,
+		0b000001101001,
+		0b000001101010,
+		0b000001101011,
+		0b000011010010,
+		0b000011010011,
+		0b000011010100,
+		0b000011010101,
+		0b000011010110,
+		0b000011010111,
+		0b000001101100,
+		0b000001101101,
+		0b000011011010,
+		0b000011011011,
+		0b000001010100,
+		0b000001010101,
+		0b000001010110,
+		0b000001010111,
+		0b000001100100,
+		0b000001100101,
+		0b000001010010,
+		0b000001010011,
+		0b000000100100,
+		0b000000110111,
+		0b000000111000,
+		0b000000100111,
+		0b000000101000,
+		0b000001011000,
+		0b000001011001,
+		0b000000101011,
+		0b000000101100,
+		0b000001011010,
+		0b000001100110,
+		0b000001100111
+	},
+	{
+		0b00110101,
+		0b000111,
+		0b0111,
+		0b1000,
+		0b1011,
+		0b1100,
+		0b1110,
+		0b1111,
+		0b10011,
+		0b10100,
+		0b00111,
+		0b01000,
+		0b001000,
+		0b000011,
+		0b110100,
+		0b110101,
+		0b101010,
+		0b101011,
+		0b0100111,
+		0b0001100,
+		0b0001000,
+		0b0010111,
+		0b0000011,
+		0b0000100,
+		0b0101000,
+		0b0101011,
+		0b0010011,
+		0b0100100,
+		0b0011000,
+		0b00000010,
+		0b00000011,
+		0b00011010,
+		0b00011011,
+		0b00010010,
+		0b00010011,
+		0b00010100,
+		0b00010101,
+		0b00010110,
+		0b00010111,
+		0b00101000,
+		0b00101001,
+		0b00101010,
+		0b00101011,
+		0b00101100,
+		0b00101101,
+		0b00000100,
+		0b00000101,
+		0b00001010,
+		0b00001011,
+		0b01010010,
+		0b01010011,
+		0b01010100,
+		0b01010101,
+		0b00100100,
+		0b00100101,
+		0b01011000,
+		0b01011001,
+		0b01011010,
+		0b01011011,
+		0b01001010,
+		0b01001011,
+		0b00110010,
+		0b00110011,
+		0b00110100
+	}
+};
+
+static const unsigned t4_term_length[2][64] = {
+	{10, 2, 2, 3, 3, 4, 4, 5, 6, 6, 7, 7, 7, 8, 8, 9, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12},
+	{8, 6, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8}
+};
+
+static const unsigned t4_min_term_length[] = {2, 4};
+static const unsigned t4_min_make_up_length[] = {10, 5};
+
+static const unsigned t4_max_term_length[] = {12, 8};
+static const unsigned t4_max_make_up_length[] = {13, 9};
+
+static const unsigned t4_make_up_length[2][15] = {
+	{10, 12, 12, 12, 12, 12, 12, 13, 13, 13, 13, 13, 13, 13, 13},
+	{5, 5, 6, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9}
+};
+
+static const unsigned t4_make_up_ind[15] = {64, 128, 192, 256, 320, 384, 448, 512, 576, 640, 704, 768, 832, 896, 960};
+
+static const unsigned t4_make_up[2][15] = {
+	{
+		0b0000001111,
+		0b000011001000,
+		0b000011001001,
+		0b000001011011,
+		0b000000110011,
+		0b000000110100,
+		0b000000110101,
+		0b0000001101100,
+		0b0000001101101,
+		0b0000001001010,
+		0b0000001001011,
+		0b0000001001100,
+		0b0000001001101,
+		0b0000001110010,
+		0b0000001110011
+	},
+	{
+		0b11011,
+		0b10010,
+		0b010111,
+		0b0110111,
+		0b00110110,
+		0b00110111,
+		0b01100100,
+		0b01100101,
+		0b01101000,
+		0b01100111,
+		0b011001100,
+		0b011001101,
+		0b011010010,
+		0b011010011,
+		0b011010100
+	 }
+};
+
+/*! \brief Attempt to decode compressed bit vector
+ *
+ * Return length of RLE according to modified ITU-T T.4 from TS 44.060 Table 9.1.10.2
+ * or -1 if no applicable RLE found
+ * N. B: we need explicit bit length to make decoding unambiguous
+*/
+static inline int t4_rle_term(unsigned w, bool b, unsigned bits)
+{
+	unsigned i;
+	for (i = 0; i < 64; i++)
+		if (w == t4_term[b][i] && bits == t4_term_length[b][i])
+			return i;
+	return -1;
+}
+
+static inline int t4_rle_makeup(unsigned w, bool b, unsigned bits)
+{
+	unsigned i;
+	for (i = 0; i < 15; i++)
+		if (w == t4_make_up[b][i] && bits == t4_make_up_length[b][i])
+			return t4_make_up_ind[i];
+	return -1;
+}
+
+/*! \brief Make-up codes for a given length
+ *
+ * Return proper make-up code word for an uninterrupted sequence of b bits
+ * of length len according to modified ITU-T T.4 from TS 44.060 Table 9.1.10.2 */
+static inline int t4_rle(struct bitvec *bv, unsigned len, bool b)
+{
+	if (len >= 960) {
+		bitvec_set_uint(bv, t4_make_up[b][14], t4_make_up_length[b][14]);
+		return bitvec_set_uint(bv, t4_term[b][len - 960], t4_term_length[b][len - 960]);
+	}
+
+	if (len >= 896) {
+		bitvec_set_uint(bv, t4_make_up[b][13], t4_make_up_length[b][13]);
+		return bitvec_set_uint(bv, t4_term[b][len - 896], t4_term_length[b][len - 896]);
+	}
+
+	if (len >= 832) {
+		bitvec_set_uint(bv, t4_make_up[b][12], t4_make_up_length[b][12]);
+		return bitvec_set_uint(bv, t4_term[b][len - 832], t4_term_length[b][len - 832]);
+	}
+
+	if (len >= 768) {
+		bitvec_set_uint(bv, t4_make_up[b][11], t4_make_up_length[b][11]);
+		return bitvec_set_uint(bv, t4_term[b][len - 768], t4_term_length[b][len - 768]);
+	}
+
+	if (len >= 704) {
+		bitvec_set_uint(bv, t4_make_up[b][10], t4_make_up_length[b][10]);
+		return bitvec_set_uint(bv, t4_term[b][len - 704], t4_term_length[b][len - 704]);
+	}
+
+	if (len >= 640) {
+		bitvec_set_uint(bv, t4_make_up[b][9], t4_make_up_length[b][9]);
+		return bitvec_set_uint(bv, t4_term[b][len - 640], t4_term_length[b][len - 640]);
+	}
+
+	if (len >= 576) {
+		bitvec_set_uint(bv, t4_make_up[b][8], t4_make_up_length[b][8]);
+		return bitvec_set_uint(bv, t4_term[b][len - 576], t4_term_length[b][len - 576]);
+	}
+
+	if (len >= 512) {
+		bitvec_set_uint(bv, t4_make_up[b][7], t4_make_up_length[b][7]);
+		return bitvec_set_uint(bv, t4_term[b][len - 512], t4_term_length[b][len - 512]);
+	}
+
+	if (len >= 448) {
+		bitvec_set_uint(bv, t4_make_up[b][6], t4_make_up_length[b][6]);
+		return bitvec_set_uint(bv, t4_term[b][len - 448], t4_term_length[b][len - 448]);
+	}
+
+	if (len >= 384) {
+		bitvec_set_uint(bv, t4_make_up[b][5], t4_make_up_length[b][5]);
+		return bitvec_set_uint(bv, t4_term[b][len - 384], t4_term_length[b][len - 384]);
+	}
+
+	if (len >= 320) {
+		bitvec_set_uint(bv, t4_make_up[b][4], t4_make_up_length[b][4]);
+		return bitvec_set_uint(bv, t4_term[b][len - 320], t4_term_length[b][len - 320]);
+	}
+
+	if (len >= 256) {
+		bitvec_set_uint(bv, t4_make_up[b][3], t4_make_up_length[b][3]);
+		return bitvec_set_uint(bv, t4_term[b][len - 256], t4_term_length[b][len - 256]);
+	}
+
+	if (len >= 192) {
+		bitvec_set_uint(bv, t4_make_up[b][2], t4_make_up_length[b][2]);
+		return bitvec_set_uint(bv, t4_term[b][len - 192], t4_term_length[b][len - 192]);
+	}
+
+	if (len >= 128) {
+		bitvec_set_uint(bv, t4_make_up[b][1], t4_make_up_length[b][1]);
+		return bitvec_set_uint(bv, t4_term[b][len - 128], t4_term_length[b][len - 128]);
+	}
+
+	if (len >= 64) {
+		bitvec_set_uint(bv, t4_make_up[b][0], t4_make_up_length[b][0]);
+		return bitvec_set_uint(bv, t4_term[b][len - 64], t4_term_length[b][len - 64]);
+	}
+
+	return bitvec_set_uint(bv, t4_term[b][len], t4_term_length[b][len]);
+}
+
+enum dec_state {
+	EXPECT_TERM,
+	TOO_LONG,
+	NEED_MORE_BITS,
+	CORRUPT,
+	OK
+};
+
+static inline enum dec_state _t4_step(struct bitvec *v, uint16_t w, bool b, unsigned bits, bool term_only)
+{
+	if (bits > t4_max_make_up_length[b])
+		return TOO_LONG;
+	if (bits < t4_min_term_length[b])
+		return NEED_MORE_BITS;
+
+	if (term_only) {
+		if (bits > t4_max_term_length[b])
+			return CORRUPT;
+		int t = t4_rle_term(w, b, bits);
+		if (-1 != t) {
+			bitvec_fill(v, t, b ? ONE : ZERO);
+			return OK;
+		}
+		return NEED_MORE_BITS;
+	}
+
+	int m = t4_rle_makeup(w, b, bits);
+	if (-1 != m) {
+		bitvec_fill(v, m, b ? ONE : ZERO);
+		return EXPECT_TERM;
+	}
+
+	m = t4_rle_term(w, b, bits);
+	if (-1 != m) {
+		bitvec_fill(v, m, b ? ONE : ZERO);
+		return OK;
+	}
+
+	return NEED_MORE_BITS;
+}
+
+/*! \brief decode T4-encoded bit vector
+ *  Assumes MSB first encoding.
+ *  \param[in] in bit vector with encoded data
+ *  \param[in] cc color code (whether decoding should start with 1 or 0)
+ *  \param[out] out the bit vector to store result into
+ * returns 0 on success, negative value otherwise
+ */
+int osmo_t4_decode(const struct bitvec *in, bool cc, struct bitvec *out)
+{
+	uint8_t orig[in->data_len];
+	struct bitvec vec;
+	vec.data = orig;
+	vec.data_len = in->data_len;
+	bitvec_zero(&vec);
+	memcpy(vec.data, in->data, in->data_len);
+	vec.cur_bit = in->cur_bit;
+
+	/* init decoder using known color code: */
+	unsigned bits = t4_min_term_length[cc];
+	enum dec_state d;
+	int16_t w = bitvec_get_int16_msb(&vec, bits);
+	bool b = cc;
+	bool term_only = false;
+
+	while (vec.cur_bit > 0) {
+		d = _t4_step(out, w, b, bits, term_only);
+
+		switch (d) {
+		case EXPECT_TERM:
+			bitvec_shiftl(&vec, bits);
+			bits = t4_min_term_length[b];
+			w = bitvec_get_int16_msb(&vec, bits);
+			term_only = true;
+			break;
+		case OK:
+			bitvec_shiftl(&vec, bits);
+			bits = t4_min_term_length[!b];
+			w = bitvec_get_int16_msb(&vec, bits);
+			b = !b;
+			term_only = false;
+			break;
+		case NEED_MORE_BITS:
+			bits++;
+			w = bitvec_get_int16_msb(&vec, bits);
+			break;
+		case TOO_LONG:
+			return -E2BIG;
+		case CORRUPT:
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+/*! \brief encode bit vector in-place using T4 encoding
+ *  Assumes MSB first encoding.
+ *  \param[in] bv bit vector to be encoded
+ * returns color code (if the encoding started with 0 or 1) or -1 on failure (encoded is bigger than original)
+ */
+int osmo_t4_encode(struct bitvec *bv)
+{
+	unsigned rl0 = bitvec_rl(bv, false), rl1 = bitvec_rl(bv, true);
+	int r = (rl0 > rl1) ? 0 : 1;
+	uint8_t orig[bv->data_len], tmp[bv->data_len * 2]; /* FIXME: better estimate max possible encoding overhead */
+	struct bitvec comp, vec;
+	comp.data = tmp;
+	comp.data_len = bv->data_len * 2;
+	bitvec_zero(&comp);
+	vec.data = orig;
+	vec.data_len = bv->data_len;
+	bitvec_zero(&vec);
+	memcpy(vec.data, bv->data, bv->data_len);
+	vec.cur_bit = bv->cur_bit;
+
+	while (vec.cur_bit > 0) {
+		if (rl0 > rl1) {
+			bitvec_shiftl(&vec, rl0);
+			t4_rle(&comp, rl0, false);
+		} else {
+			bitvec_shiftl(&vec, rl1);
+			t4_rle(&comp, rl1, true);
+		}
+		/*
+		  TODO: implement backtracking for optimal encoding
+		  printf(" -> [%d/%d]", comp.cur_bit + vec.cur_bit, bv->cur_bit);
+		*/
+		rl0 = bitvec_rl(&vec, false);
+		rl1 = bitvec_rl(&vec, true);
+	}
+	if (comp.cur_bit < bv->cur_bit) {
+		memcpy(bv->data, tmp, bv->data_len);
+		bv->cur_bit = comp.cur_bit;
+		return r;
+	}
+	return -1;
+}
+
+