CSV module cannot handle the case of embedded \r (i.e.
carriage return) in a field.

As far as I can see, this is hard-coded into the _csv.c
file and cannot be fixed with Dialect changes.

Logged In: YES 
user_id=80475

Skip, does this coincide with your planned switchover to
universal newlines?

Logged In: YES 
user_id=698599

I suspect this restriction (CR appearing within a quoted 
field) is a historical accident and can be safely removed. 

Logged In: YES 
user_id=44345

It certainly intersects with it somehow.  ;-)  If nothing else, it
will serve as a useful test case.

Logged In: YES 
user_id=698599

If you're interested, I've just checked in a change to the CVS head for 
Python 2.5 that may, at least partially, fix this problem (if you try it, let me 
know how it goes).

Logged In: YES 
user_id=7733

I just filed a bug (http://www.python.org/sf/1465014) that
seems to be related to this. Revision 38290 on
Modules/_csv.c includes the addition of this code:

    else if (c == '\n' || c == '\r') {
  	self->state = EAT_CRNL;
  	break;
    }

(and similar). This seems to be eating (deleting) control
chars, but newlines used to be significant. 

Embedded line breaks are allowed, according to RFC 4180
(http://www.ietf.org/rfc/rfc4180.txt). And according to the
Wikipedia entry
(http://en.wikipedia.org/wiki/Comma-separated_values), "a
line break within an element must be preserved."

IIUC, I get the correct behavior:

trunk-py$ ./python ~/Desktop/tcsv.py
['fld1', 'fld2', 'fld3 ', 'fld4']
['fld1', 'fld2', 'fld3 \r', 'fld4']

trunk-py$ cat ~/Desktop/tcsv.py
#! /usr/local/bin/python

import csv

d = 'fld1,fld2,"fld3 ",fld4\r\n'
d2 = 'fld1,fld2,"fld3 \r'
d3 = '",fld4\r\n'

r = csv.reader([d, d2, d3], dialect="excel")
for f in r:
        print f