[idn] ACE Candidates: Dropping MACE in favor of AMC-ACE-Z

[Makoto Ishisone <[e-mail]>]

We, the authors of MACE, which is one of the proposed ACEs, would like
to annouce that we now support AMCE-ACE-Z over MACE as the ACE
standard.

Currently there are 3 ACE candidates in Pool W -- DUDE, AMC-ACE-Z and
MACE.

DUDE has the simplest algorithm, but we feel that it is not very
efficient, especially for CJK scripts.  Typically it allows 15 Han
characters (in the worst case 14) as the report from the ACE design
team says (draft-ietf-idn-ace-report-00), but many people seem to be
feeling 15 is insufficient.  Also this figure is worse than 17 that is
RACE's worst case in CJK.  We think choosing ACE that is not only
simpler but also more efficient than RACE is important.

That's why we proposed MACE.  We focused on improving the worst case
behavior especially for CJK.  It is designed to allow at least 19 Han
(or Hangul) characters, but it is slightly worse compression for
Western scripts. Also it has additional complexity by the introduction
of "encoding modes."

AMC-ACE-Z uses a brand new compression algorithm, and has clearly the
better compression for Western scripts compared to DUDE/MACE.  And we
found it also has the efficiency comparable to MACE for CJK (at least
Japanese) names.  We have done an evaluation using 50,000 registered
Japanese domain names, and it showed that AMC-ACE-Z does better
compression on average than MACE, and slightly worse compression for
worst case (but by only one or two octets, which is still better than
RACE).  Also AMC-ACE-Z's algorithm is simpler than MACE.

So, considering all the factors, we decided to support AMC-ACE-Z
rather than revising/enhancing MACE.

						Makoto Ishisone
						[e-mail]
						Yoshiro Yoneya
						[e-mail]