SC22/WG20 N796

From: Kenneth Whistler [kenw@sybase.com]
Sent: Monday, October 23, 2000 6:38 PM
CLAUI TD Sophia-Antipolis Recommendations
(concerns SC2, SC22, SC32, SC35 and SC7)

Alain,

For the record, I find myself disagreeing with all or parts of
the Sophia recommendations #7, #8, and #9:

Sophia 7. I don't think that a type-3 TR on "implementation,
interoperability and migration guidelines for coded character
sets" is a useful production out of JTC1. The danger, depending
on who tries to develop it, is that it is too likely to be a
rehash of the hash that TC304 made in its Character Encoding
Guidelines document. Also, the whole concept of a TR "initiated
by SC2 with joint SC22/WG20 editorship" is unwieldy in the
extreme.

The SC2 recommendation and guideline can be simply stated:

"Transition all IT applications to ISO/IEC 10646, as quickly as
feasible."

And that is exactly what industry is attempting to do.

As to the difficulties that end users have making things work during
the transition, as they deal with products that aren't all
seamlessly integrated, that is best dealt with by information
on websites (generic and company-specific), FAQ's, how-to books,
magazine articles, technical chatrooms with experienced people
providing hints and tips to newbies, etc.

I fail to see how having an ISO TR on this topic would make even
a slight dent in the end-user confusion and need for detailed
how-to hints on working with particular products (that are
continually changing, by the way).

Sophia 8. I don't think it is a good idea at all for JTC1 to be
encouraging developers of other international standards to be
making normative references to type-1 TRs (failed standard). The
very fact that a standard failed to achieve consensus should
be a big red flag of danger. As we see from DTR 14652, once
a failed standard goes to type-1 TR status, it may also
fail to achieve consensus in its content, but may roll ahead
to approval, with little detailed technical oversight, simply
because those involved no longer care about it enough to make
sure it is "right". Pushing this off to another group to make
normative references to bad material, simply because they can
reach consensus to do so (perhaps in the absence of any detailed
technical analysis of what they are referring to), is a recipe
for bad standard making. This is clearly yet another effort to
raise the status of 14652 in SC22, once again.

Sophia 9. The liaison part of this recommendation is fine. TC46/SC2
shouldn't be making transliteration standards without reference
to 10646 at this point. However, the recommendation to do this in
cooperation with SC22/WG20 "which already has a specification format
(in DTR 14652) to describe transliteration between different languages,"
is another attempt to misappropriate a bad specification in a failed
standard, to do more than it should. The DTR 14652 specification for
transliteration is ambiguous and unsophisticated. A much more
worked-out approach can be found in the ICU classes, for example.

--Ken