The new draft correctly states how to hash the user-id to make
a signature, but it left out some padding data:
Werner (finger gcrypt(_at_)ftp(_dot_)guug(_dot_)de for
info about GnuPG)
Werner is correct. In the case of V4 signature packets, 6 more bytes
are added at the end, for technical reasons to prevent some signature
aliasing attacks. The contents are:
byte 0 - Version of signature packet, in this case 4
byte 1 - 0xff
bytes 2-5 - Length of hashed data from sig packet (not including
these last 6 bytes), bigendian
We added the hashing information to the draft at the final stages, and
it was an oversight on my part in not getting this information to Jon
correctly.
The information on V3 signature hashing is correct in the draft.
Hal Finney