on the state of PGP compatibility (2nd try)

[Adam Back]

[This is actually slightly more accurate and even worse than my first
mail which bounced to some of the lists as I had a typo, _and_
separately encountered a mail hub outage at cyberpass.net -- apologies
to those who get duplicates].

So I was trying to decrypt this stored mail sent to me by a GPG user,
and lo pgp6.x failed to decrypt it.

So I try an older gpg I had installed, and it fails because it doesn't
support RSA or IDEA, and this GPG user it seems installed the RSA and
IDEA patches.

So I go fetch GPG from www.gnupg.org, but it still doesn't contain
IDEA by default due to the anti-patent religion thing gone-to-far over
at GNU land, so I try to download the idea.c plugin -- except they
seem to have removed it (accidentally? hmm -- I fire off a mildly
brusque email to webmaster at gnupg.org -- and hit google.com but all
the first few hits are pointing at the gnupg.org faq with labyrinth of
links ending eventually in the same dud link.)

Either way _even_ if idea.c were still where they claimed it would be
it seems intentionally very well hidden, which I think is
unconscionable for a security product: to _intentionally_ frustrate
users attempts to interoperate with secure previous versions and
alternate implementations.

So then I try pgp5.x but the binary is using dynamic libraries that
are no longer on my shiny new redhat7.1 installation, so I try to
compile it but it no longer compiles.  Tinker briefly fixing up
things, but the errors are multiple, and I haven't got time for this.

So my last hope is pgp2.x, but some buggy pgp variant has left my
pgp2.x key ring empty, and you can't use it directly on pgp6.x
keyrings as it will barf on the new key formats.  So then I ponder
exporting my private key out of pgp6.5.8 which isn't going to do it
compatibly without some serious thought -- openPGP's salted key
stretching maybe being used -- do I want to export it without a
password (not really), so figure out how do I turn the salted key
stretching off, will pgp6.5.8 even let you export private keys, or is
it easier to just extract it with a binary editor?  Fortunately I
finally find a pgp2.x keyring secring.bak file (thanks PRZ), and move
it back and lo pgp2 can't decrypt it because of some unsupported
packetry.  I take a look at it with Mark Shoulsen's pgpacket and it
seems that _even_ pgpacket thinks there is some unsupported packets at
the end, so dig out another packet analysis program -- pgpdump by Kazu
Yamamoto and it doesn't seem to realise it's ascii armored or is
expecting to find an external program to de-armor which is missing --
I don't care, so use emacs and mmencode -u to produce a binary
version, and then it plays.  And there lies the problem: gpg encrypted
it with IDEA using the new openPGP streaming options to encode the
message in chunks despite it being encrypted with idea (presumably the
sender forget to invoke --rfc1991 not realising my potential future
predicament).  Thus sprach Kazu's analyzer:

New: Symmetrically Encrypted Data Packet(tag 9)(512 bytes) partial start
	Encrypted data [if pub/sym session key not present, sym alg - IDEA]
New:	(201 bytes) partial end

So, for now, give up.  I guess it's cheaper to just send the original
author an email ask him if he remembers that idea he sent me 4 months
ago and have him send me it in clear text to be sure!

What a nightmare!  Try that sequence on a novice user and they give up
before they get past the first GPG faq with rant about algorithm
patents.

We've really got to do something about the compatibility problems.

Adam
--
http://www.cypherspace.org/adam/

---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at wasabisystems.com