Re: [openpgp] Marker packet for OpenPGP-NG (was: Confirming open questions discussed at IETF 114)

["Neal H. Walfield" <[e-mail]>]

On Wed, 12 Oct 2022 17:00:40 +0200,
Werner Koch wrote:
> 1. All data including key packets shall make use of a marker packet
> 
>    5.8.  Marker Packet (Tag 10)
> 
>    This packet was used by certain PGP versions and flagged as obsolete
>    in RFC-4880.  With this specification it is repurposed to mark the
>    data and keys constructed as specified by this specification and its
>    successors.
> 
>    The body of this packet consists of:
> 
>      - The three octets 0x60, 0x67, 0x70  (which spell "pgp" in UTF-8).

(I think you mean 0x70, 0x67, 0x70.)

>    Implementations SHALL use this packet to indicate packet formats
>    described by this specification.  It SHALL be the first packet of
>    each OpenPGP Message or sets of OpenPGP Keys.  Note that the old
>    marker packet used three different octets.

I agree with this suggestion, but for different reasons.

Currently, fingerprinting a PGP message is tricky.  Consider the
file(1) command / libmagic(3) library.  libmagic implements a domain
specific language (magic) to fingerprint files.  To confidently (>= 48
bits of entropy) fingerprint binary OpenPGP keys, I wrote a relatively
complicated parser:

  https://github.com/file/file/blob/master/magic/Magdir/pgp-binary-keys

Compare this with, say, the ogg audio format, which is identified in
just one line of magic:

  https://github.com/file/file/blob/master/magic/Magdir/vorbis#L25

(The rest of the file extracts some meta-data.)  Basically, the file
contains a fixed string ("OggS") at a fixed location (offset 0).

I think it would be good if OpenPGP also had such a string.  And, the
marker packet is a pretty good way to add it in a non-intrusive way, I
think.

Neal