PEP 456: drop pluggable and go for compile time configuration of the hash algorithm.

pep-0456.txt

@@ -1,5 +1,5 @@
 PEP: 456
-Title: Pluggable and secure hash algorithm
+Title: Secure and interchangeable hash algorithm
 Version: $Revision$
 Last-Modified: $Date$
 Author: Christian Heimes <christian@python.org>
@@ -8,16 +8,16 @@ Type: Standards Track
 Content-Type: text/x-rst
 Created: 27-Sep-2013
 Python-Version: 3.4
-Post-History:
+Post-History: 06-Oct-2013
 
 
 Abstract
 ========
 
 This PEP proposes SipHash as default string and bytes hash algorithm to properly
-fix hash randomization once and for all. It also proposes an addition to
+fix hash randomization once and for all. It also proposes modifications to
-Python's C API in order to make the hash code pluggable. The new API allows to
+Python's C code in order to unify the hash code and to make it easily
-select the algorithm on startup as well as the addition of more hash algorithms.
+interchangeable.
 
 
 Rationale
@@ -57,10 +57,8 @@ This PEP proposes three major changes to the hash code for strings and bytes:
   ``Objects/object.c`` and ``Objects/unicodeobject.c``. The function takes a
   void pointer plus length and returns the hash for it.
 
-* The algorithm can be selected by the user with an environment variable,
+* The algorithm can be selected at compile time. FNV is guaranteed to exist
-  command line argument or with an API function (for embedders). FNV is
+  on all platforms. SipHash is available on the majority of modern systems.
-  guaranteed to exist on all platforms. SipHash is available on the majority
-  of modern systems.
 
 
 Requirements for a hash function
@@ -321,50 +319,25 @@ hash function table
 type definition::
 
     typedef struct {
-        PyHash_Func hashfunc;  /* function pointer */
+        PyHash_Func hash;      /* function pointer */
         char *name;            /* name of the hash algorithm and variant */
         int hash_bits;         /* internal size of hash value */
         int seed_bits;         /* size of seed input */
-        int precedence;        /* ranking for auto-selection */
+    } _PyHash_FuncDef;
-    } PyHash_FuncDef;
 
-    PyAPI_DATA(PyHash_FuncDef *) PyHash_FuncTable;
+    PyAPI_DATA(_PyHash_FuncDef *) _PyHash_Func;
 
 Implementation::
 
-    PyHash_FuncDef hash_func_table[] = {
+    #ifndef PY_HASH_FUNC
-        {fnv, "fnv", 64, 128, 10},
     #ifdef PY_UINT64_T
-        {siphash24, "sip24", sizeof(Py_hash_t)*8, sizeof(Py_hash_t)*8, 20},
+    _PyHash_Func = {siphash24, "sip24", 64, 128}
+    #else
+    _PyHash_Func = {fnv, "fnv", 8 * sizeof(Py_hash_t), 16 * sizeof(Py_hash_t)}
+    #endif
     #endif
-        {NULL, NULL},
-    };
 
-    PyHash_FuncDef *PyHash_FuncTable = hash_func_table;
+TODO: select hash algorithm with autoconf variable
-
-
-hash function API
------------------
-
-function proto types::
-
-    PyAPI_FUNC(int) PyHash_SetHashAlgorithm(char *name);
-
-    PyAPI_FUNC(PyHash_FuncDef *) PyHash_GetHashAlgorithm(void);
-
-    PyAPI_DATA(PyHash_FuncDef *) _PyHash_Func;
-
-``PyHash_SetHashAlgorithm(NULL)`` selects the hash algorithm with the highest
-precedence. ``PyHash_SetHashAlgorithm("sip24")`` selects siphash24 as hash
-algorithm. The function returns ``0`` on success. In case the algorithm is
-not supported or a hash algorithm is already set it returns ``-1``.
-(XXX use enum?)
-
-``PyHash_GetHashAlgorithm()`` returns a pointer to current hash function
-definition or `NULL`.
-
-``_PyHash_Func`` holds the set hash function definition. It can't be modified
-or reset once a hash algorithm is set.
 
 
 Python API addition
@@ -379,9 +352,8 @@ algorithm as well as all available algorithms.
 ::
 
     sys.hash_info(algorithm='siphash24',
-                  available_algorithms=('siphash24', 'fnv'),
                   hash_bits=64,
-                  hash_output=64,        # sizeof(Py_hash_t)*8
+                  hash_output=64,        # 8 * sizeof(Py_hash_t)
                   seed_bits=128)
 
 
@@ -439,7 +411,7 @@ multiplied with the size of the internal unicode kind::
 
     if (PyUnicode_READY(u) == -1)
         return -1;
-    x = _PyHash_Func->hashfunc(PyUnicode_DATA(u),
+    x = _PyHash_Func->hash(PyUnicode_DATA(u),
                            PyUnicode_GET_LENGTH(u) * PyUnicode_KIND(u));
 
 
@@ -534,6 +506,19 @@ the past, but are not subject of this PEP.
    prefixes are stored within the tree structure.
 
 
+Discussion
+==========
+
+Pluggable
+---------
+
+The first draft of this PEP made the hash algorithm pluggable at runtime. It
+supported multiple hash algorithms in one binary to give the user the
+possibility to select a hash algorithm at startup. The approach was considered
+an unnecessary complication by several core committers [pluggable]_. Subsequent
+versions of the PEP aim for compile time configuration.
+
+
 Reference
 =========
 
@@ -567,6 +552,8 @@ Reference
 
 .. [aes-ni] http://en.wikipedia.org/wiki/AES_instruction_set
 
+.. [pluggable] https://mail.python.org/pipermail/python-dev/2013-October/129138.html
+
 
 Copyright
 =========