Re: PGP/GPG and AOL

[Simon Josefsson <jas@extundo.com>]

On Tue, 27 Nov 2001, Jack Twilley wrote:

> Jack> I helped an AOL user install WinPT on her machine, but when she
> Jack> sent me an encrypted message, Gnus and gpg.el didn't DTRT.
> 
> Simon> Could you give some more details on what Gnus/gpg.el did and
> Simon> did not do?
> 
> What Gnus and gpg.el did not do was to take the content of either part
> of the multipart message and offer to decrypt it and display the
> decrypted contents.

What happened instead?  What was displayed?  Did `K b' help?

> Jack> Turns out that AOL sends multipart messages -- text and HTML --
> Jack> and there's no known way to force it to send only text.
> 
> Simon> Shouldn't be a problem.
> 
> You'd think.  I've got an open case with AOL because first and second
> level helpdesk folks weren't able to tell me how to do that.

Oh, I meant it shouldn't be a problem for Gnus.  To get AOL to do 
something is probably more difficult, yes. :-)

> 
> Jack> Does anyone out here have an easy way to have Gnus and gpg.el
> Jack> check the contents of a multipart message to see if they're
> Jack> encrypted, and offer to decrypt them?
> 
> Simon> Gnus should do this by default, controlled by
> Simon> `mm-decrypt-option'.
> 
> Mine's set to nil, which means "ask user" according to the docs.  I
> was not asked.

Then Gnus didn't regard it as encrypted, which could be a bug.

The best would be if you could arrange to send a encrypted mail from the
AOL account to bugs@gnus.org and I could try to see what is wrong.

> While the "right" solution is to fix AOL, the more probable solution
> seems to be to fix Gnus or gpg.el -- does anyone know what's wrong or
> how to fix it?  I'll help debug the problem if people can give me
> generic pointers on just what to do, as I'm unfamiliar with this
> particular subsystem.

The generic hint is to use elp-instrument-package to find out which 
functions are called when you display the message, and then edebug some of 
likely functions to find out what is happening. (Functions named *crypt*, 
*secur* are good candidates, if you can't find any, just edebug everything 
in mm-*.el.)   It takes some time to get to know the code though.