Re: [Evolution] Preliminary list of feature requests [Long]

[Nathan Thompson-Amato <ndt jps net>]

Jeffrey Stedfast wrote:

> decryption was is already working and I've started working on encryption
> today. I've got GnuPG encryption working but I haven't started PGP5 and
> PGP2 yet. Also, there's no GUI hooks to allow encryption but as soon as
> there is, it should be easy to have outgoing messages encrypted (and
> optionally signed).

Ack!  I started work on this a few hours ago.  I should've posted to the
list first, I guess. :-)

> I'm guessing that we should also have an interface to just pgp sign an
> outgoing message? I also assume verify would be important...

Yes, definitely.  These are harder to do interface-wise, though -- see
Jamie Zawinski's rant a few weeks ago on evolution-hackers.

> I'm also not quite clear as to the best GUI way for doing PGP encryption,
> signing, and verification anyone have suggestions?

IMHO, checkboxes will do for the first two -- I'm picturing the same
interface as the "send HTML mail" option, with a global default and a
per-message override.

The complications happen when you send an encrypted message (a) to
somebody who doesn't have a public key, and (b) to several people, some
of which don't have public keys.  One idea is to offer to grab public
keys from a keyserver whenever you add a key-less recipient to the
message and have the "encrypt" box checked.

For the verification UI, I'll just quote Jamie:

> We spent literally *months* arguing about this w.r.t. the S/MIME support
> in Netscape.  What we ended up with was putting a crypto-stamp thingy up
> in the message header area, to the right of the table that contains
> From/To/Subject/etc.  That's an area that can't be touched by the HTML
> in the message body (in the case of top-level messages -- you could of
> course send someone an HTML message that looked like a forwarded message
> that had been signed.)
>
> But I think it might be the case that if the message body plays games
> with layers, it can overwrite that area.  I'm not sure.  I think someone
> might have put in a hack to prevent that (some magic internal tag beyond
> which a layer couldn't render or something.)
>
> Basically, it's a hard problem and probably not worth worrying about too
> much: if it's a message that is important enough that someone would go
> to the trouble of spoofing the signature, then it's probably a message
> where the recipient actually cares about the signature, and will click
> on the link to see details about that signature instead of just trusting
> the single bit of information ("yes/no") that the UI displays.

Nathan