From: gnu(_at_)toad(_dot_)com
Half of why PGP is now up and running all over the world, while PEM is
still stumbling out of the starting gate, is that PGP completely eliminates
two bottlenecks built into the PEM design:
* you can't pick your own name
* you have to register with an authority
I tried to use PEM, but TIS would not give me the name I requested
(O=gnu(_at_)cygnus(_dot_)com).
I have to agree the I am very umimpressed with all this X.500 complexity
and huffing and puffing about "sovereign states" controlling naming, etc.
X.500 strict hierarchial paralysis around countries is simply what you get
from a political process rather than a techncial one (not that it is ever
completely black or white).
PEM's model is that somehow the naming conventions that everyone is
already using (user(_at_)do(_dot_)main) are ridiculous or inappropriate -- so
let's invent a new kind of naming (DN's), and then introduce
translations between them at every user. (E.g. I send mail to
cerf(_at_)vint(_dot_)net, it gets translated locally to c=xxx,o=yyy,foo=bar,
then
the key for that c=xxx,o=yyy,foo=bar is looked up locally, then the
message is encrypted and sent. Why that first translation?)
My model is that Internet domain email works *JUST FINE* for me, while
every piece of mail I receive via an X.nnn gateway is full of
screwiness. (E.g. MCI Mail from Esther Dyson now contains a
140-character return address.) I want a direct mapping between
ordinary email names and names in PEM certificates.
<enter sarcasm mode> I particularly like the habit many X.400 gateways
have of either throwing away my mail without telling me or sometimes
sending a bounce indication back with no hint of what message it was
that bounced, other than the addressee. (no date, subject line, body copy,
etc.) <exit sarcasm mode>
Someone suggested C=US,O=gnu(_at_)cygnus(_dot_)com(_dot_) But cygnus.com is
valid in any
country. I could move to Australia and keep gnu(_at_)cygnus(_dot_)com(_dot_)
It's not tied
to geography. The Domain Name System is up and running worldwide. Let's
use it.
Well, who knows, maybe there are people with that sort of name somewhere...
but I don't see any problems at all with
C=INTERNET,O=user(_at_)f(_dot_)q(_dot_)d(_dot_)n(_dot_) Someone
should start a service issuing PEM certificates like and if the PEM hierarchy
won't let they joint, they should just start their own hierarchy.
Of course, there is no problem with duplicate assignments, since there
is already a multi level registration of domain names and user names.
Happily *that* system doesn't care what name you register -- it only
rejects names if they are already in use.
The main objection raised to this sort of easy and obvious name is,
"When the X.500 revolution comes, your name will be lined up against
the wall and shot". I'm perfectly willing to take that chance, given
my own personal estimate of the usefulness of X.400 and X.500.
Let me guess -- about twenty influential people on this list are
unwilling for me to *have* that choice. And that's why PGP is
winning, and will continue to win, though it is inferior technically.
Because it doesn't impose arbitrary and inappropriate models on its
users: it just encrypts, decrypts, hashes, certifies, looks up keys,
and does key maintenance. Funny, that. It might even end up
integrated into MIME before PEM does.
I waiting for the next round of PEM/MIME integration drafts to study
and make sure they are usable with PGP. In fact, you should be able
to send a message with both PEM and PGP and X-whatever signatures. If
the proposed format won't allow that, it needs to be expanded to be
able to do so.
John Gilmore
Donald E. Eastlake, III