On August 6, 2005 at 20:12, wayne wrote:
Others have pointed out the weakness of the "nowsp"
I too have wondered why the canonicalization algorigthm proposed in DK
(and now DKIM) eleminates all white space instead of collasping white
space.
Pardon my ignorance, but can someone tell me why?
Eric Allman provided some of the reasoning behind nowsp:
<http://www.mhonarc.org/archive/cgi-bin/mesg.cgi?a=ietf-mailsig&i=999F4AD48066042BA78AEB53%40irma.local>
I'm not sure the "one space" proposal solves all problems, but, unless
I'm missing something, it seems like a better start.
I think the "one space" algorithm may have unattended consequences
for some text media types (e.g. text/tab-separated-values). Ignoring
the argument of how often other IANA registered text media-types
are actually used, message body issues with whitespace tend to deal
with trailing whitespace at the end of lines and the end of the body.
Whitespace at the beginning of the message body may also be of
a concern (as noted in past messages to the list).
I initially proposed an alternative to nowsp in my first comments
about the DKIM draft:
<http://www.mhonarc.org/archive/cgi-bin/mesg.cgi?a=ietf-mailsig&i=200507142321.j6ENL3B03008%40gator.earlhood.com>
And then further refined it (calling it minwsp) after further list
discussion on the topic:
<http://www.mhonarc.org/archive/cgi-bin/mesg.cgi?a=ietf-mailsig&i=200507182154.j6ILsLZ09567%40gator.earlhood.com>
The minwsp algoritm is less liberal than my initial proposal, but
we may not care to support software that changes whitespace characters
(e.g. SP -> TAB). My first proposal did the whitespace compression
technique that onesp employs, but I limited it to header fields.
--ewh