chinese language search doesn't work, redux.

I've noticed the documented fact that search in Chinese is don't work
out of the box.

Before submitting this, I searched around this site and found a few
references to this problem. The most up to date seemed to be:
http://drupal.org/node/2142
in which Steven suggests that you use a break-word preprocessor
hook_search_preprocess on chinese text to split the text into "words".

The way it's described in 2142, "Chinese doesn't use spaces". But I
think a better characterization is "Chinese doesn't have words". I've
read a research paper (I think on automated word segmentation by Dekai
Wu) which mentioned that chinese native speakers themselves have a
pretty large disagreement of where to put "word breaks" in Chinese
text, and that presumably comes from the fact that they don't
necessarily think of everything in terms of "words." I know that it's
never clear to me (as a non-native non-literate sort-of-speaker) which
things might be considered "words". A lot of particles could be words
by themselves, or maybe attached to the verb or noun that they modify.
A lot of nouns are compounds, whose component parts make sense as
words by themselves.

So what's the point? I suggest that:

Out of the box, index all chinese characters as if they were
separate words.

Here's the reasoning:

1. This is close to the ideal behavior anyway. A search for 我 should find ALL text that contains that character. A search for 我的 should ideally find all text that contains the string "我的", but an okay second place would be to find all text that contained both 我 and 的. Later, a feature could be added to rank search results based on the order of keywords in the original search query, so that "我的" results came up before results with 我 and 的 in different parts of the page.
2. It seems that to implement that behavior at the moment, I need to reduce the "index words less than this length" parameter of the search indexer to "1," and add a hook_search_preprocess hook to do something like add a space before and after every cjk character. That's unideal for the english content on the site.
3. What I'd like is to not have to reduce the minimum word length to index to "1." I guess I'd still have to reduce the minimum word length to search for to "1."
4. This seems like it could be done by building the default chinese word splitting into the search indexing engine, since you'd want to index all Chinese "words" even though they were all 1 character long.
5. Indexing all Chinese characters shouldn't be a big deal. Most sites probably won't have any Chinese characters in their content. For the sites that do, consider that the vast majority of chinese text draws from a bag of the most commonly used 6000-10,000 characters. That's a much better soft upper bound than the number of words you might encounter on english language pages.
6. It's far better than the current out of the box behavior, which doesn't work at all.

Thoughts?