I have been looking over the PEM I-Ds and find myself wanting to re-open
an old argument. Please indulge this message.
In the absence of historical reasons, I am at a loss to explain why PEM
doesn't use MIME. All you need to is define a single content-type, e.g.,
message/pem
which is formatted as an rfc-822 message. The headers of this message
are the usual PEM headers plus some Content- headers to identify the
body which follows (if the PEM headers indicate CRL, then there is no
body and the Content- headers are absent). In most cases, there would
be two Content- headers:
Content-Type: message/rfc822
Content-Transfer-Encoding: base64
One argument made against this is that PEM would become dependent on
MIME, and that the size of the PEM community would be limited to being a
subset of the MIME community. My response is that there is nothing
which prevents PEM software from implementing precisely the minimal
amount of MIME needed for the above support. This equates to an
encapsulation scheme which is a bit different than the one PEM current
uses, and the use of base64 which PEM already uses. In brief: it
doesn't make it any harder to write a PEM UA, and later down the road it
would make it simpler because PEM implementors could leverage off MIME
implementations. A counter-response, I guess, is that this would impact
the few existing PEM implementations.
I want to stress here that in no way does this scheme lock in the PEM
community to being fully MIME-capable. It means that PEM user agents
will generate encapsulations in a format which is consistent with the
way MIME does it, and that there will be a couple of extra headers added.
So, at the risk of being accused of poisoning the well, let me ask,
other than the "we don't want to change because PEM will be published as an
RFC tomorrow" argument, is there some other argument against this proposal?
Thanks for your patience,
/mtr
ps: In case it isn't obvious, Stef and I are the authors of rfc934 which
is the encapsulation scheme currently used by PEM. I have given up on
rfc934 because the scheme in MIME is superior. There is at least one
degenerate case in which repeated encapsulations of a message may result
in lossage due to implementation restrictions of some message transfer
agents. Whether or not these implementations are broken is immaterial,
what must be emphasized is that breakage will occur!