Re: encrypt.el in No Gnus 0.7

[Ted Zlatanov <tzz@lifelogs.com>]

On Mon, 05 Nov 2007 21:16:18 -0500 Richard Stallman <rms@gnu.org> wrote: 

RS>     Emacs is built around the idea of letting users do what they want, isn't
RS>     it?

RS> In Emacs, users can change any Lisp code at run time.  This general
RS> facility, which lets users do absolutely whatever they want, means we
RS> do not need to provide specific customization features for everything
RS> somebody would like to change.  We only provide specific customization
RS> features where they are substantially useful.

RS> Providing every hook that someone might imaginably want is an explicit
RS> non-goal of Emacs development.  (And always has been.)

All right.  I know what I find useful, and EasyPG is too limited in what
it does to be useful for general encryption.  It does a particular job
well, which is interfacing with GnuPG, but I think it's a disservice to
Emacs users to assume that GnuPG is all they need, and that they want to
install it to get encryption/decryption services.  If you disagree with
me, there's little more I can say to convince you otherwise, since this
is the essential point I'm trying to make.

RS>     The package was in Gnus, and I don't have bug reports.  I have not
RS>     advertised the library so far, except for brief notes to Gnus users.  So
RS>     I don't have a sample big enough to answer your question.

RS> In other words, there is no indication that users actually use this
RS> particular customization facility.

As I explained, I have not advertised it.  You're playing catch-22,
justifying to yourself that no one will use the library with the lack of
users before it's been advertised.  To put it in your language, there is
also no indication that users don't use encrypt.el.  The sample size is
too small.

RS> You could post a request where Gnus users will see it, asking them to
RS> let you know if they have used it.

By now they have, this thread is CCed to the Gnus mailing list.

RS>     I found a few GPL ones (just a sample list) with a simple web search:

RS>     MCrypt http://mcrypt.sourceforge.net/
RS>     AxCrypt http://www.axantum.com/AxCrypt/
RS>     ScramDisk 4 Linux http://sourceforge.net/projects/sd4l/
RS>     Several steganography tools (hide data inside an image, for example)

RS> We are miscommunicating.  I think you mean that in principle they
RS> could be useful for encryption.  Maybe they are.  But the question is
RS> whether this particular feature is useful, not whether other useful
RS> encryption programs exist.

Steganography is useful, in particular, and GnuPG doesn't support it.
Users could hide secret information, or messages for other users, or
watermarks.  This data can hide inside images or other binary data in a
public directory.  This seems obviously useful to me.  Maybe you're
considering just encryption, rather than general cryptography (of which
steganography is generally considered a part).  We could change the
package name to cipher.el or ecode.el if that makes more sense to you.

RS>     - pure Lisp ciphers for quick experimentation and sharing

RS> That seems unimportant to me.

I think cryptography is a field where experimentation is very
important.  Wouldn't you like to see Emacs become a breeding ground for
cryptography work?  I would love to see that.  GnuPG is not a good
vehicle for experimentation, since it requires C code and lots of prep
work to do even simple things (as I understand it from reading the
docs).

RS> The main use of simple obfuscation ciphers is in sending mail, and for
RS> that purpose, you cannot tell from a file name (there's none) or a
RS> buffer name (it's the same as always) whether to do it.

Obfuscation can be used for a lot more, but I don't know if that
argument is all that interesting.  Steganography is probably the more
convincing argument, since it's essentially obfuscation by
concealment, and there's plenty of interest in it.

Ted