Indexing multiple numbers in one field

Before I learned all the details of how Plucene works, I devised an 
indexing scheme for a simple Web application, an application that allows 
multiple users to take, save, index and share notes. The notes live in a 
relational database, but I would use Plucene for text indexing. My text 
indexing schema looked roughly like this:

note / UnStored (search default)
title / Text
created / Text
updated / Text
user_ids / Text

...where user_ids consisted of multiple numbers separated by spaces. I 
need to track user_ids so that Sally doesn't see Janes notes except in 
cases where Jane has set up a special sharing relationship. I 
used Plucene::Plugin::Analyzer::PorterAnalyzer as the analyzer so I would 
get stemming for the notes.

Of course, this does not work. First of all, 
Plucene::Analysis::LetterTokenizer defines a token regex of 
/[[:alpha:]]+/, which does not match numbers. Secondly, 
Lingua::Stem::En::stem deletes anything that is not [A-Za-z].

I can define user_ids to be a Plucene::Document::Field of type Keyword, 
which works fine if there is only one user_id. But this fails to handle 
multiple user_ids.

I am curious for thoughts on the best course of action.

I can change user_ids to user_id (of type Keyword) and simply enter 
an instance of the document into the index for each associated user.

I can write my own analyzer, but this will entail replacing/editing not 
only LetterTokenizer but also Lingue::Stem::En. (Nevertheless, I'm 
leaning in this direction. I'd like people to be able to run a search 
like "Windows 3.0" or "perl5" and get a meaningful result.)

I can write a new field class, called Keywords, which would be just like 
Keyword except tokenized based on whitespace or some other optional field 
separator.

Or I could use a separate index for each user, but I have ruled out this 
approach because eventually some notes will be shared among quite a few 
users.

Thoughts?

Cheers
Ryan