diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..6350e98
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+.coverage
diff --git a/bin/swh-hashfile b/bin/swh-hashfile
new file mode 100755
index 0000000..af066e5
--- /dev/null
+++ b/bin/swh-hashfile
@@ -0,0 +1,15 @@
+#!/usr/bin/python3
+
+# Copyright (C) 2015  The Software Heritage developers
+# See the AUTHORS file at the top-level directory of this distribution
+# License: GNU General Public License version 3, or any later version
+# See top-level LICENSE file for more information
+
+import sys
+
+from swh.core.hashutil import hashfile
+
+if __name__ == '__main__':
+    fname = sys.argv[1]
+    for (algo, checksum) in sorted(hashfile(fname).items()):
+        print('%s\t%s' % (algo, checksum))
diff --git a/swh/__init__.py b/swh/__init__.py
new file mode 100644
index 0000000..3ad9513
--- /dev/null
+++ b/swh/__init__.py
@@ -0,0 +1,2 @@
+from pkgutil import extend_path
+__path__ = extend_path(__path__, __name__)
diff --git a/swh/core/__init__.py b/swh/core/__init__.py
new file mode 100644
index 0000000..fdffa2a
--- /dev/null
+++ b/swh/core/__init__.py
@@ -0,0 +1 @@
+# placeholder
diff --git a/swh/core/hashutil.py b/swh/core/hashutil.py
new file mode 100644
index 0000000..08bd0dc
--- /dev/null
+++ b/swh/core/hashutil.py
@@ -0,0 +1,88 @@
+# Copyright (C) 2015  The Software Heritage developers
+# See the AUTHORS file at the top-level directory of this distribution
+# License: GNU General Public License version 3, or any later version
+# See top-level LICENSE file for more information
+
+import hashlib
+import os
+
+from io import BytesIO
+
+# supported hashing algorithms
+ALGORITHMS = set(['sha1', 'sha256', 'sha1_git'])
+
+# should be a multiple of 64 (sha1/sha256's block size)
+# FWIW coreutils' sha1sum uses 32768
+HASH_BLOCK_SIZE = 32768
+
+
+def _new_hash(algo, length=None):
+    """Initialize a digest object (as returned by python's hashli) for the
+    requested algorithm. See the constant ALGORITHMS for the list of supported
+    algorithms. If a git-specific hashing algorithm is requested (e.g.,
+    "sha1_git"), the hasing object will be pre-fed with the needed header; for
+    this to work, length must be given.
+
+    """
+    if algo not in ALGORITHMS:
+        raise ValueError('unknown hashing algorithm ' + algo)
+
+    h = None
+    if algo.endswith('_git'):
+        if length is None:
+            raise ValueError('missing length for git hashing algorithm')
+        h = hashlib.new(algo.split('_')[0])
+        h.update(('blob %d\0' % length).encode('ascii'))  # git hash header
+    else:
+        h = hashlib.new(algo)
+
+    return h
+
+
+def _hash_file_obj(f, length, algorithms=ALGORITHMS):
+    """hash the content of a file-like object
+
+    """
+    hashers = {algo: _new_hash(algo, length)
+               for algo in algorithms}
+    while True:
+        chunk = f.read(HASH_BLOCK_SIZE)
+        if not chunk:
+            break
+        for h in hashers.values():
+            h.update(chunk)
+
+    return {algo: hashers[algo].hexdigest() for algo in hashers}
+
+
+def _hash_fname(fname, algorithms=ALGORITHMS):
+    """hash the content of a file specified by file name
+
+    """
+    length = os.path.getsize(fname)
+    with open(fname, 'rb') as f:
+        return _hash_file_obj(f, length)
+
+
+def hashfile(f, length=None, algorithms=ALGORITHMS):
+    """Hash the content of a given file, given either as a file-like object or a
+    file name. All specified hash algorithms will be computed, reading the file
+    only once. Returns a dictionary mapping algorithm names to hex-encoded
+    checksums.
+
+    When passing a file-like object, content length must be given; when passing
+    a file name, content length is ignored.
+
+    """
+    if isinstance(f, str):
+        return _hash_fname(f, algorithms)
+    else:
+        return _hash_file_obj(f, length, algorithms)
+
+
+def hashdata(data, algorithms=ALGORITHMS):
+    """Like hashfile, but hashes content passed as a string (of bytes)
+
+    """
+    buf = BytesIO(data)
+    return _hash_file_obj(buf, len(data), algorithms)
diff --git a/swh/tests/test_hashlib.py b/swh/tests/test_hashlib.py
new file mode 100644
index 0000000..21b416f
--- /dev/null
+++ b/swh/tests/test_hashlib.py
@@ -0,0 +1,48 @@
+# Copyright (C) 2015  The Software Heritage developers
+# See the AUTHORS file at the top-level directory of this distribution
+# License: GNU General Public License version 3, or any later version
+# See top-level LICENSE file for more information
+
+import tempfile
+import unittest
+
+from nose.tools import istest
+
+from swh.core import hashutil
+
+
+class Hashlib(unittest.TestCase):
+
+    def setUp(self):
+        self.data = b'42\n'
+        self.checksums = {
+            'sha1':     '34973274ccef6ab4dfaaf86599792fa9c3fe4689',
+            'sha1_git': 'd81cc0710eb6cf9efd5b920a8453e1e07157b6cd',
+            'sha256':   '084c799cd551dd1d8d5c5f9a5d593b2e931f5e36122ee5c793c1d08a19839cc0',  # NOQA
+            }
+
+    @istest
+    def hashdata(self):
+        checksums = hashutil.hashdata(self.data)
+        self.assertEqual(checksums, self.checksums)
+
+    @istest
+    def unknown_algo(self):
+        with self.assertRaises(ValueError):
+            hashutil.hashdata(self.data, algorithms=['does-not-exist'])
+
+    @istest
+    def hashfile_by_name(self):
+        with tempfile.NamedTemporaryFile() as f:
+            f.write(self.data)
+            f.flush()
+            checksums = hashutil.hashfile(f.name)
+            self.assertEqual(checksums, self.checksums)
+
+    @istest
+    def hashfile_by_obj(self):
+        with tempfile.TemporaryFile() as f:
+            f.write(self.data)
+            f.seek(0)
+            checksums = hashutil.hashfile(f, len(self.data))
+            self.assertEqual(checksums, self.checksums)