I didn't ignore it, I'm pretty sure we discussed it.
Your reply didn't show up in my mailbox nor in the archive at
www.imc.org, then.
Now, an *implementation* of OpenPGP has to be always generating thingies
trimmed. When it does, they will interoperate with the old versions just
fine. The only subtle gotcha here is that when *receiving* a signed text
object, a smart implementation won't throw a hissy fit if the text object
hasn't been trimmed; it will just roll its eyes and mutter "get with the
program, geezer" under its breath. This is the old principle of "be liberal
in what you accept and conservative in what you generate."
That sort of makes sense, but the RFC should have contained a notice
about the incompatibility. When you follow the letter of the RFC, the
obvious thing to do when you verify signatures is to use the same
function for tweaking the line ends for both signature types.
Following the principle of "being liberal in what you accept" without
knowing about this specific incompatibility (there are dozens of ways
a signing implementation could fail) is impossible because there is no
such thing as a "near miss" with a message digest.