On Thu, Jan 05, 2006 at 05:52:40PM +0100, Daniel A. Nagy wrote:
I wrote:
In order to
prevent arbitrary wraps, the length of actual lines should be maximized in
64 characters
You wrote:
This would give a valid encoding for long header lines, but does not
fix the problem at hand. Word wrapping would still result in broken
armor as word wrapping wouldn't know to put whitespace before the
lines it reformats.
There would be no word wrapping, if everything I propose becomes standard.
In common usage, a user creating an armored message (or clearsigned
message which has the same problem) uses an OpenPGP program and some
editor. These two programs do not communicate, so I do not see how
the OpenPGP program can know where to fold the lines to prevent the
editor from coming along and wrapping. Whatever (reasonable) magic
number is used as the maximum number of columns, I am confident there
is a word wrap setting that will clobber it.
This problem basically exists in the "Comment" and "Version" header
lines. Both are essentially comments, are optional, and carry nothing
vital to OpenPGP. Rather than invent a whole new encoding for long
lines, that involves modifying many deployed parsers for the benefit
of having longer comments, why not just add a warning to the draft
that says in effect "long comments may break you - be careful".
David