Noteworthy changes in version 0.4.2 ----------------------------------- * Fixed this huge memory leak. * Redesigned the trust database: You should run "gpgm --check-trustdb". New command --update-trustdb, which adds new key from the public keyring into your trustdb * Fixed a bug in the armor code, leading to invalid packet errors. (a workaround for this was to use --no-armor). * comment packets are not anymore generated. "--export" filters them out. One Exception: The comment packets in a secret keyring are still used because they carry the factorization of the public prime product. * --import now only looks for KEYBLOCK headers, so you can now simply remove the "- " in front of such a header if someone accdently signed such a message or the keyblock is part of a cleartext signed message. * --with-colons now lists the key expiration time and not anymore the valid period. Noteworthy changes in version 0.4.1 ----------------------------------- * A secondary key is used when the primary key is specified but cannot be used for the operation (if it is a sign-only key). * GNUPG can now handle concatenated armored messages: There is still a bug if different kinds of messages are mixed. * Iterated+Salted passphrases now work. If want to be sure that PGP5 is able to handle them you may want to use the options "--s2k-mode 3 --s2k-cipher-algo cast5 --s2k-digest-algo sha1" when changing a passphrase. * doc/OpenPGP talks about OpenPGP compliance, doc/HACKING gives a few hints about the internal structure. * Checked gnupg against the August 1998 draft (07) and I believe it is in compliance with this document (except for one point). * Fixed some bugs in the import merging code and rewrote some code for the trustdb. Noteworthy changes in version 0.4.0 ----------------------------------- * Triple DES is now supported. Michael Roth did this piece of needed work. We have now all the coded needed to be OpenPGP compliant. * Added a simple rpm spec file (see INSTALL). * detached and armored signatures are now using "PGP SIGNATURE", except when --rfc1991 is used. * All times which are not in the yyy-mm-dd format are now printed in local time. Noteworthy changes in version 0.3.5 ----------------------------------- * New option --throw-keyid to create anonymous enciphered messages. If gpg detects such a message it tires all available secret keys in turn so decode it. This is a gnupg extension and not in OpenPGP but it has been discussed there and afaik some products use this scheme too (Suggested by Nimrod Zimmerman). * Fixed a bug with 5 byte length headers. * --delete-[secret-]key is now also availabe in gpgm. * cleartext signatures are not anymore converted to LF only. * Fixed a trustdb problem. Run "gpgm --check-trustdb" to fix old trust dbs. * Building in another directory should now work. * Weak key detection mechanism (Niklas Hernaeus). Noteworthy changes in version 0.3.4 ----------------------------------- * New options --comment and --set-filename; see g10/OPTIONS * yes/no, y/n localized. * Fixed some bugs. Noteworthy changes in version 0.3.3 ----------------------------------- * IMPORTANT: I found yet another bug in the way the secret keys are encrypted - I did it the way pgp 2.x did it, but OpenPGP and pgp 5.x specifiy another (in some aspects simpler) method. To convert your secret keys you have to do this: 1. Build the new release but don't install it and keep a copy of the old program. 2. Disable the network, make sure that you are the only user, be sure that there are no trojan horses etc .... 3. Use your old gpg (version 0.3.[12]) and set the passphrases of ALL your secret keys to empty! (gpg --change-passphrase your-user-id). 4. Save your ownertrusts (see the next point) 5. rm ~/.gnupg/trustd.gpg 6. install the new version of gpg (0.3.3) 7. For every secret key call "gpg --edit-key your-user-id", enter "passwd" at the prompt, follow the instructions and change your passward back, enter "save" to store it. 8. Restore the ownertrust (see next point). * The format of the trust database has changed; you must delete the old one, so gnupg can create a new one. IMPORTANT: Use version 0.3.[12] to save your assigned ownertrusts ("gpgm --list-ownertrust >saved-trust"); then build this new version and restore the ownertrust with this new version ("gpgm --import-ownertrust saved-trust"). Please note that --list-ownertrust has been renamed to --export-ownertrust in this release and it does now only export defined ownertrusts. * The command --edit-key now provides a commandline driven menu which can be used vor vaious tasks. --sign-key is only an an alias to --edit-key and maybe removed in future: use the command "sign" of this new menu - you can select which user ids you want to sign. * Alternate user ids can now be created an signed. * Owner trust values can now be changed with --edit-key (trust) * GNUPG can now run as a coprocess; this enables sophisticated frontends. tools/shmtest.c is a simple sample implemenation. This needs some more work: all tty_xxx() are to be replaced by cpr_xxx() and some changes in the display logics is needed. * Removed options --gen-prime and --gen-random. * Removed option --add-key; use --edit-key instead. * Removed option --change-passphrase; use --edit-key instead. * Signatures are now checked even if the output file could not be created. Command "--verify" tries to find the detached data. * gpg now disables core dumps. * compress and symmetric cipher preferences are now used. Because there is no 3DES yet, this is replaced by Blowfish. * We have added the Twofish as an experimental cipher algorithm. Many thanks to Matthew Skala for doing this work. Twofish is the AES submission from Schneier et al.; see "www.counterpane.com/twofish.html" for more information. * Started with a help system: If you enter a question mark at some prompt; you should get a specific help for this prompt. * There is no more backup copy of the secret keyring. * A lot of new bugs. I think this release is not as stable as the previous one. Noteworthy changes in version 0.3.2 ----------------------------------- * Fixed some bugs when using --textmode (-seat) * Now displays the trust status of a positive verified message. * Keyrings are now scanned in the sequence they are added with --[secret-]keyring. Note that the default keyring is implictly added as the very first one unless --no-default-keyring is used. * Fixed setuid and dlopen bug. Noteworthy changes in version 0.3.1 ----------------------------------- * Partial headers are now written in the OpenPGP format if a key in a v4 packet is used. * Removed some unused options, removed the gnupg.sig stuff. * Key lookup by name now returns a key which can be used for the desired action. * New options --list-ownertrust (gpgm) to make a backup copy of the ownertrust values you assigned. * clear signature headers are now in compliance with OpenPGP. Noteworthy changes in version 0.3.0 ----------------------------------- * New option --emulate-checksum-bug. If your passphrase does not work anymore, use this option and --change-passphrase to rewrite your passphrase. * More complete v4 key support: Preferences and expiration time is set into the self signature. * Key generation defaults to DSA/ElGamal keys, so that new keys are interoperable with pgp5 * DSA key generation is faster and key generation does not anymore remove entropy from the random generator (the primes are public parameters, so there is really no need for a cryptographic secure prime number generator which we had used). * A complete new structure for representing the key parameters. * Removed most public key knowledge into the cipher libray. * Support for dynamic loading of new algorithms. * Moved tiger to an extension module. Noteworthy changes in version 0.2.19 ------------------------------------ * Replaced /dev/urandom in checks with new tool mk-tdata. * Some assembler file cleanups; some more functions for the Alpha. * Tiger has now the OpenPGP assigned number 6. Because the OID has changed, old signatures using this algorithm can't be verified. * gnupg now encrypts the compressed packed and not any longer in the reverse order; anyway it can decrypt both versions. Thanks to Tom for telling me this (not security related) bug. * --add-key works and you are now able to generate subkeys. * It is now possible to generate ElGamal keys in v4 packets to create valid OpenPGP keys. * Some new features for better integration into MUAs. Noteworthy changes in version 0.2.18 ------------------------------------ * Splitted cipher/random.c, add new option "--disable-dev-random" to configure to support the development of a random source for other systems. Prepared sourcefiles rand-unix.c, rand-w32.c and rand-dummy.c (which is used to allow compilation on systems without a random source). * Fixed a small bug in the key generation (it was possible that 48 bits of a key were not taken from the random pool) * Add key generation for DSA and v4 signatures. * Add a function trap_unaligned(), so that a SIGBUS is issued on Alphas and not the slow emulation code is used. And success: rmd160 raised a SIGBUS. * Enhanced the formatting facility of argparse and changed the use of \r,\v to @ because gettext does not like it. * New option "--compress-algo 1" to allow the creation of compressed messages which are readable by PGP and "--print-md" (gpgm) to make speed measurement easier. Noteworthy changes in version 0.2.17 ------------------------------------ * Comment packets are now of private type 61. * Passphrase code still used a 160 bit blowfish key, added a silly workaround. Please change your passphrase again - sorry. * Conventional encryption now uses a type 3 packet to describe the used algorithms. * The new algorithm number for Blowfish is 20, 16 is still used for encryption only; for signing it is only used when it is in a v3 packet, so that GNUPG keys are still valid. Noteworthy changes in version 0.2.16 ------------------------------------ * Add experimental support for the TIGER/192 message diigest algorithm. (But there is only a dummy ASN OID). * Standard cipher is now Blowfish with 128 bit key in OpenPGP's CFB mode. I renamed the old cipher to Blowfish160. Because the OpenPGP group refused to assign me a number for Blowfish160, I have to drop support for this in the future. You should use "--change-passphrase" to recode your current passphrase with 128 bit Blowfish. Noteworthy changes in version 0.2.15 ------------------------------------ * Fixed a bug with the old checksum calculation for secret keys. If you run the program without --batch, a warning does inform you if your secret key needs to be converted; simply use --change-passphrase to recalculate the checksum. Please do this soon, as the compatible mode will be removed sometime in the future. * CAST5 works (using the PGP's special CFB mode). * Again somewhat more PGP 5 compatible. * Some new test cases Noteworthy changes in version 0.2.14 ------------------------------------ * Changed the internal handling of keyrings. * Add support to list PGP 5 keyrings with supkeys * Timestamps of signatures are now verified. * A expiration time can now be specified during key generation. * Some speedups for Blowfish and SHA-1, rewrote SHA-1 transform. Reduced the amount of random bytes needed for key generation in some cases. Noteworthy changes in version 0.2.13 ------------------------------------ * Verify of DSA signatures works. * Re-implemented the slower random number generator. Noteworthy changes in version 0.2.12 ------------------------------------ * --delete-key checks that there is no secret key. The new option --delete-secret-key maybe used to delete a secret key. * "-kv" now works as expected. Options "--list-{keys,sigs]" and "--check-sigs" are now working. * New options "--verify" and "--decrypt" to better support integration into MUAs (partly done for Mutt). * New option "--with-colons" to make parsing of key lists easier. Noteworthy changes in version 0.2.11 ------------------------------------ * GPG now asks for a recipient's name if option "-r" is not used. * If there is no good trust path, the program asks whether to use the public keys anyway. * "--delete-key" works for public keys. What semantics shall I use when there is a secret key too? Delete the secret key or leave him and auto-regenerate the public key, netxt time the secret key is used? Noteworthy changes in version 0.2.10 ------------------------------------ * Code for the alpha is much faster (about 20 times); the data was misaligned and the kernel traps this, so nearly all time was used by system to trap the misalignments and to write syslog messages. Shame on me and thanks to Ralph for pointing me at this while drinking some beer yesterday. * Changed some configure options and add an option --disable-m-guard to remove the memory checking code and to compile everthing with optimization on. * New environment variable GNUPGHOME, which can be used to set another homedir than ~/.gnupg. Changed default homedir for Windoze version to c:/gnupg. * Fixed detached signatures; detached PGP signatures caused a SEGV. * The Windoze version works (as usual w/o a strong RNG). Noteworthy changes in version 0.2.9 ----------------------------------- * Fixed FreeBSD bug. * Added a simple man page. * Switched to automake1.2f and a newer gettext. Noteworthy changes in version 0.2.8 ----------------------------------- * Changed the name to GNUPG, the binaries are called gpg and gpgm. You must rename rename the directory "~/.g10" to ~/.gnupg/, rename {pub,sec}ring.g10 to {pub,sec}ring.gpg, trustdb.g10 to trustdb.gpg and g10.sig to gnupg.sig. * New or changed passphrases are now salted. Noteworthy changes in version 0.2.7 ----------------------------------- * New command "gen-revoke" to create a key revocation certificate. * New option "homedir" to set the homedir (which defaults to "~/.g10"). This directory is created if it does not exists (only the last part of the name and not the complete hierarchy) * Command "import" works. (Try: "finger gcrypt@ftp.guug.de|g10 --import") * New commands "dearmor/enarmor" for g10maint. These are mainly used for internal test purposes. * Option --version now conforming to the GNU standards and lists the available ciphers, message digests and public key algorithms. * Assembler code for m68k (not tested). * "make check" works. Noteworthy changes in version 0.2.6 ----------------------------------- * Option "--export" works. Noteworthy changes in version 0.2.5 ----------------------------------- * Added zlib for systems which don't have it. Use "./configure --with-zlib" to link with the static version. * Generalized some more functions and rewrote the encoding of message digests into MPIs. * Enhanced the checkit script Noteworthy changes in version 0.2.4 ----------------------------------- * nearly doubled the speed of the ElGamal signature verification. * backup copies of keyrings are created. * assembler stuff for Pentium; gives about 15% better perfomance. * fixed a lot of bugs. Noteworthy changes in version 0.2.3 ----------------------------------- * Found a bug in the calculation of ELG fingerprints. This is now fixed, but all existing fingerprints and keyids for ELG keys are not any more valid. * armor should now work; including clear signed text. * moved some options to the new program g10maint * It's now 64 bit clean and runs fine on an alpha--linux. * Key generation is much faster now. I fixed this by using not so strong random number for the primes (this was a bug because the ElGamal primes are public parameters and it does not make sense to generate them from strong random). The real secret is the x value which is still generated from strong (okay: /dev/random) random bits. * added option "--status-fd": see g10/OPTIONS * We have secure memeory on systems which support mlock(). It is not complete yet, because we do not have signal handler which does a cleanup in very case. We should also check the ulimit for the user in the case that the admin does not have set a limit on locked pages. * started with internationalization support. * The logic to handle the web of trust is now implemented. It is has some bugs; but I'm going to change the algorithm anyway. It works by calculating the trustlevel on the fly. It may ask you to provide trust parameters if the calculated trust probability is too low. I will write a paper which discusses this new approach. * a couple of changes to the configure script. * New option "--quick-random" which uses a much quicker random number generator. Keys generated while this option is in effect are flags with "INSECURE!" in the user-id. This is a development only option. * Read support for new version packets (OpenPGP). * Comment packets are now of correct OpenPGP type 16. Old comment packets written by G10 are detected because they always start with a hash which is an invalid version byte. * The string "(INSECURE!)" is appended to a new user-id if this is generated on a system without a good random number generator.