From: Charles Lindsey (chl@clw.cs.man.ac.uk)
Date: Tue Dec 03 2002 - 08:49:33 CST
Votes were received from 18 members of the Working Group.
Issue 1 - (8BIT HEADERS)
------------------------
#X CONSENSUS LIMIT-----------------------------------------------------------+
#6 ALL 8bit ENCODED; NO I18N GROUPS---------------------------------------+ |
#5 ALL 8bit ENCODED on the wire----------------------------------------+ | |
#4 SEND 8bit REGARDLESS---------------------------------------------+ | | |
#3 8bit in MODERATED - SHOULD NOT YET----------------------------+ | | | |
#2 8bit in MODERATED - MUST NOT-------------------------------+ | | | | |
#1 STRICT EMAIL COMPLIANCE---------------------------------+ | | | | | |
#0 the PRESENT DRAFT------------------------------------+ | | | | | | |
| | | | | | | |
Bill McQuillan 6 3 5 4 8 1 2 7
Claus Färber 6 5 8 7 4 1 2 3
Kent Landfield 2 1 4 3 5 6 7 4
Thorfinn 2 1 3 4 5 7 8 6
Russ Allbery 7 3 8 6 4 1 2 5
Charles Lindsey 2 3 6 1 4 7 8 5
Per Abrahamsen 4 3 1 2 6 5 8 7
Paul Overell 5 2 3 4 7 1 6 2
Forrest J. Cavalier 3 3 4 4 5 1 1 2
Erland Sommarsko 1 4 3 2 5 8 7 6
Shmuel (Seymour J.) Metz 1 6 3 2 4 5 7 6
Clive Feather 2 2 2 2 1 4 5 3
Jean-Marc Desperrier 2 5 4 1 3 7 6 5
Henry Spencer 1 2 7 5 6 3 8 4
John Moreno 1 2 5 4 3 6 7 4
G Andruk 7 4 5 6 8 2 1 3
John Stanley - - - - - - 1 -
Ralph Babel 4 4 4 4 4 1 2 3
RESULT:
#1 beats #0 the PRESENT DRAFT by 8 : 6
#1 beats #2 8bit in MODERATED - MUST NOT by 11 : 4
#1 beats #3 8bit in MODERATED - SHOULD NOT YET by 10 : 5
#1 beats #4 SEND 8bit REGARDLESS by 12 : 4
#1 beats #5 ALL 8bit ENCODED on the wire by 9 : 8
#1 beats #6 ALL 8bit ENCODED; NO I18N GROUPS by 11 : 7
#1 beats #X CONSENSUS LIMIT by 10 : 4
#1 STRICT EMAIL COMPLIANCE is therefore the condorcet winner
The following matrix shows the votes in more detail.
The number at [row,column] indicates the number of voters who placed the
[row] option ahead of the [column] option in their order of preference
#X CONSENSUS LIMIT----------------------------------------------+
| #6 ALL 8bit ENCODED; NO I18N GROUPS-----------------------+ |
| | #5 ALL 8bit ENCODED on the wire---------------------+ | |
| | | #4 SEND 8bit REGARDLESS-----------------------+ | | |
| | | | #3 8bit in MODERATED - SHOULD NOT YET---+ | | | |
| | | | | #2 8bit in MODERATED - MUST NOT---+ | | | | |
| | | | | | #1 STRICT EMAIL COMPLIANCE--+ | | | | | |
| | | | | | | #0 the PRESENT DRAFT--+ | | | | | | |
| | | | | | | +------------------ \\ 6 11 8 13 10 11 11
| | | | | | +-------------------- 8 \\ 11 10 12 9 11 10
| | | | | +---------------------- 4 4 \\ 4 9 9 11 7
| | | | +------------------------ 7 5 10 \\ 12 9 11 9
| | | +-------------------------- 3 4 7 4 \\ 8 10 9
| | +---------------------------- 7 8 8 8 9 \\ 13 10
| +------------------------------ 7 7 7 7 8 4 \\ 7
+-------------------------------- 6 4 9 7 8 7 11 \\
The second matrix shows the effect of subtracting [column,row] from
[row,column] so as to give the majority in favour of [row] as against
[column]. Observe that the winning row(s) has no negative entries.
#X CONSENSUS LIMIT----------------------------------------------+
| #6 ALL 8bit ENCODED; NO I18N GROUPS-----------------------+ |
| | #5 ALL 8bit ENCODED on the wire---------------------+ | |
| | | #4 SEND 8bit REGARDLESS-----------------------+ | | |
| | | | #3 8bit in MODERATED - SHOULD NOT YET---+ | | | |
| | | | | #2 8bit in MODERATED - MUST NOT---+ | | | | |
| | | | | | #1 STRICT EMAIL COMPLIANCE--+ | | | | | |
| | | | | | | #0 the PRESENT DRAFT--+ | | | | | | |
| | | | | | | +------------------ \\ -2 7 1 10 3 4 5
| | | | | | +-------------------- 2 \\ 7 5 8 1 4 6 winner
| | | | | +---------------------- -7 -7 \\ -6 2 1 4 -2
| | | | +------------------------ -1 -5 6 \\ 8 1 4 2
| | | +-------------------------- -10 -8 -2 -8 \\ -1 2 1
| | +---------------------------- -3 -1 -1 -1 1 \\ 9 3
| +------------------------------ -4 -4 -4 -4 -2 -9 \\ -4
+-------------------------------- -5 -6 2 -2 -1 -3 4 \\
As expected, the Group was split into two camps; those who would accept
raw UTF-8 in some form and those who would not. Nevertheless the majority
of the Group is in favour of #1. However, we also have to look at how well
#1 beat #X, the Consensus Limit.
Now one of the interesting features of the Condorcet system is that, by
eliminating the winner and examining the matrix that is left, you can work
our what the Group as a whole would consider as its 2nd, 3rd, etc choice,
and you can also see how each of these would fare against #X. So this
gives:
Majority over #X -------------->[ Y N A]
1st #1 STRICT EMAIL COMPLIANCE 10: 4: 4
2nd #0 the PRESENT DRAFT 11: 6: 1
3rd #3 8bit in MODERATED - SHOULD NOT YET 9: 7: 2
4th #2 8bit in MODERATED - MUST NOT 7: 9: 2
5th #5 ALL 8bit ENCODED on the wire 10: 7: 1
6th #4 SEND 8bit REGARDLESS 9: 8: 1
7th #6 ALL 8bit ENCODED; NO I18N GROUPS 7:11: 0
So it is clear that #1 also had the greatest margin over #X. So the
question now to decide is whether 10:4 of people who would accept #1 is
sufficient to amount to a "rough consensus". That is complicated by the
fact that 4 people gave equal preferences for #1 and #X. So you could say
"10 people out of 18 were willing to accept the #1 decision"
or "only 4 people out of 18 were unwilling to accept the #1 decision"
I think this has to be discussed on the List, and then our Chairman has to
declare whether that amounts to rough consensus, or else he has to take
advice from the IESG.
But, in the meantime, #1 is our preferred option. Its full text is as
follows:
a) UTF-8 is allowed in headers.
b) RFC 2047/2231 is also allowed and is required for the mailed
component of posted-and-mailed.
c) Newsgroup-names are encoded (5.2.2) when posted-and-mailed or sent to
moderators, but are raw UTF-8 on the wire.
d) Include explicit requirements on how to transform headers (for
posted-and-mailed and moderators, not necessarily for other gateways)
so as to be fully compliant with the email standards (that would
include headers in body parts of multiparts, and so recursively).
e) Injecting agents at sites that choose to carry moderated I18N groups
require to be upgraded.
f) Posts to a moderated ASCII group via non-upgraded injection agents
may get munged by MTAs en route to the moderator if they contain
UTF-8 in any headers, especially if octets in the range 0x80-9f are
present (as in maybe 73% of Unicode code points). Even though the
great majority of such posts might work, if it fails once for a given
injector/moderated-group pair, it will likely fail always for that
pair. The post would still be propagated to the ASCII group(s).
Issue 2 (SPLIT THE DOCUMENT)
----------------------------
#X CONSENSUS LIMIT-----------------------------------------------------------+
#2 NO, Don't Split it-----------------------------------------------------+ |
#1 YES, Split it-------------------------------------------------------+ | |
| | |
Bill McQuillan 1 2 3
Claus Färber 1 2 3
Kent Landfield 1 - -
Thorfinn 1 2 3
Russ Allbery 1 3 2
Charles Lindsey 3 1 2
Per Abrahamsen 2 1 3
Paul Overell 2 1 1
Forrest J. Cavalier 1 2 3
Erland Sommarsko 3 1 2
Shmuel (Seymour J.) Metz 2 1 2
Clive Feather 2 1 3
Jean-Marc Desperrier 1 2 2
Henry Spencer 2 1 3
John Moreno 2 1 1
G Andruk 1 3 2
John Stanley 1 3 2
Ralph Babel 1 3 2
RESULT:
#1 YES, Split it is preferred to #2 NO, Don't Split it by 10 : 8
#1 YES, Split it is preferred to #X CONSENSUS LIMIT by 13 : 4
#1 YES, Split it is therefore the condorcet winner
The following matrix shows the votes in more detail.
The number at [row,column] indicates the number of voters who placed the
[row] option ahead of the [column] option in their order of preference
#X CONSENSUS LIMIT------------+
| #2 NO, Don't Split it---+ |
| | #1 YES, Split it--+ | |
| | +-------------- \\ 10 13
| +---------------- 8 \\ 10
+------------------ 4 4 \\
The second matrix shows the effect of subtracting [column,row] from
[row.column] so as to give the majority in favour of [row] as against [column].
Observe that the winning row(s) has no negative entries.
#X CONSENSUS LIMIT------------+
| #2 NO, Don't Split it---+ |
| | #1 YES, Split it--+ | |
| | +-------------- \\ 2 9 winner
| +---------------- -2 \\ 6
+------------------ -9 -6 \\
The majority view here was evidently to split the document. Although the
majority was not huge, enough people placed #X as their third choice to
make it clear that the majority decision should be accepted by all. In
fact, the majority against #X was 13:4 (only 1 abstention this time) and I
think that is a clear "rough consensus".
The full text is as follows:
The document to be split into two - one for Standards Track and one for
Best Current Practice. Note that this does not involve ANY technical
change (such must be proposed separately as other issues). It is not
precluded that further "Netkeeping" material could be added to the BCP
document.
Issue 4 (MIME-STYLE HEADERS)
----------------------------
A few people "wrote in" #0 and #X. This had no effect on the result.
#X CONSENSUS LIMIT-----------------------------------------------------------+
#4 Sender, Date, Message-ID, Keywords, References, Supersedes-------------+ |
#3 Sender, Date, Message-ID, Keywords, References----------------------+ | |
#2 Sender, Date, Message-ID-----------------------------------------+ | | |
#1 Sender only---------------------------------------------------+ | | | |
#0 None-------------------------------------------------------+ | | | | |
| | | | | |
Bill McQuillan - 4 3 2 1 -
Claus Färber - 4 3 2 1 -
Kent Landfield - 2 1 3 4 -
Thorfinn - 5 3 2 1 4
Russ Allbery - 4 3 2 1 -
Charles Lindsey - 1 2 3 4 -
Per Abrahamsen - 4 3 2 1 -
Paul Overell - 4 3 2 1 -
Forrest J. Cavalier - 4 3 2 1 -
Erland Sommarsko - 4 3 2 1 -
Shmuel (Seymour J.) Metz 1 2 3 4 5 -
Clive Feather - 1 1 1 1 -
Jean-Marc Desperrier 1 2 3 4 5 -
Henry Spencer - 4 3 1 2 -
John Moreno - 4 3 2 1 -
G Andruk - 4 3 2 1 -
John Stanley - - - - 1 2
Ralph Babel - 5 4 3 1 2
RESULT:
#4 beats #0 None by 16 : 2
#4 beats #1 Sender only by 13 : 4
#4 beats #2 Sender, Date, Message-ID by 13 : 4
#4 beats #3 Sender, Date, Message-ID, Keywords, References by 12 : 5
#4 beats #X CONSENSUS LIMIT by 18 : 0
#4 Sender, Date, Message-ID, Keywords, References, Supersedes is therefore
the condorcet winner
Well, it hardly needed a fancy program to spot that! I have not bothered
to show the full matrices.
The result is clear. Here is the full text:
To remove the possibility of other-parameters (aka extension-parameters)
from the following headers that are already a part of email, so that
the article format is fully compliant with current email standards.
Sender, Date, Message-ID, Keywords, References, Supersedes
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133 Web: http://www.cs.man.ac.uk/~chl
Email: chl@clw.cs.man.ac.uk Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9 Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5