X-Git-Url: http://www.git.cypherpunks.ru/?p=pyderasn.git;a=blobdiff_plain;f=doc%2Ffeatures.rst;fp=doc%2Ffeatures.rst;h=34fe9419d3b6f4aa30dcc9f19a52c6843d68570c;hp=19cf7f97fa6ebd300c55b1f30ddc22ef28be3e7e;hb=13ea772db7870683065ac15197888cd037df78f0;hpb=abaaaec8b7a936d9a1f263d4ecc8be38f454cd73
diff --git a/doc/features.rst b/doc/features.rst
index 19cf7f9..34fe941 100644
--- a/doc/features.rst
+++ b/doc/features.rst
@@ -14,31 +14,69 @@ Features
Why yet another library? `pyasn1 `__
had all of this a long time ago. PyDERASN resembles it in many ways. In
practice it should be relatively easy to convert ``pyasn1``'s code to
-``pyderasn``'s one. But additionally it offers:
+``pyderasn``'s one.
+Also there is `asn1crypto `__.
* Small, simple and trying to be reviewable code. Just a single file
-* Automatic decoding of :ref:`DEFINED BY ` fields
+ with `six `__ dependency
* Ability to know :ref:`exact decoded ` objects offsets and
lengths inside the binary
-* :ref:`Pretty printer ` and command-line decoder, that could
- conveniently replace utilities like either ``dumpasn1`` or
- ``openssl asn1parse``
+* Automatic decoding of :ref:`DEFINED BY ` fields
+* Ability to know exact decoded field presence, emptiness: for example
+ ``SEQUENCE`` can lack ``OPTIONAL SEQUENCE OF`` field, but also can
+ have it with no elements inside
+* **Strict** DER-encoding checks. If whole input binary is parsed, then
+ it must be completely valid DER-encoded structure
+* Ability to allow BER-encoded data with knowing if any of specified
+ field has either DER or BER encoding (or possibly indefinite-length
+ encoding). For example
+ `CMS `__
+ structures allow BER encoding for the whole message, except for
+ ``SignedAttributes`` -- you can easily verify your CMS satisfies that
+ requirement
+* Extensive and comprehensive
+ `hypothesis `__
+ driven tests coverage. It also has been fuzzed with
+ `python-afl `__.
* Some kind of strong typing: SEQUENCEs require the exact **type** of
- settable values, even when they are inherited
-* However they do not require tags matching: IMPLICIT/EXPLICIT tags will
- be set automatically in the given sequence
+ settable values, even when they are inherited (assigning ``Integer``
+ to the field with the type ``CMSVersion(Integer)`` is not allowed)
+* However they do not require exact tags matching: IMPLICIT/EXPLICIT
+ tags will be set automatically in the given sequence (assigning of
+ ``CMSVersion()`` object to the field ``CMSVersion(expl=...)`` will
+ automatically set required tags)
* Descriptive errors, like ``pyderasn.DecodeError: UTCTime
(tbsCertificate:validity:notAfter:utcTime) (at 328) invalid UTCTime format``
* ``__slots__`` friendliness
-* Could be significantly faster. For example parsing of CACert.org's CRL
- under Python 3.5.2:
+* :ref:`Pretty printer ` and
+ :ref:`command-line decoder `, that could
+ conveniently replace utilities like either ``dumpasn1`` or
+ ``openssl asn1parse``
+* Could be significantly faster and have lower memory usage.
+ For example parsing of CACert.org's CRL (8.48 MiB) on FreeBSD 12.0
+ amd64, Intel Core i5-6200U 2.3 GHz machine, Python 3.5.5/2.7.15:
+
+ .. list-table::
+ :widths: 15 45 20 20
+ :header-rows: 1
- :``python -m pyderasn revoke.crl``:
- ~2 min (``pyderasn == 1.0``)
- :``python -m pyderasn --schema path.to.CertificateList revoke.crl``:
- ~38 sec (``pyderasn == 1.0``)
- :``pyasn1.decode(asn1Spec=pyasn1.CertificateList())``:
- ~22 min (``pyasn1 == 0.2.3``)
+ * - Library
+ - Command
+ - Time, sec (Py3/Py2)
+ - Memory used, MiB (Py3/Py2)
+ * - pyasn1 0.4.5
+ - ``der_decode(data, asn1Spec=rfc5280.CertificateList())``
+ - 1257 / 1302
+ - 1327 / 2093
+ * - asn1crypto 0.24.0
+ - ``asn1crypto.crl.CertificateList.load(data).native``
+ - 29.3 / 43.8
+ - 983 / 1677
+ * - pyderasn 4.9
+ - ``CertificateList().decode(data)`` (CertificateList is
+ converted ``pyasn1`` scheme definition)
+ - 27.6 / 32.5
+ - 498 / 488
There are drawbacks: