Ascii and EBCDIC translation

We do a lot of import and export of data, plus have both PC client (local
and web) input as well at PC5250.
Had a recent thread involving cut and paste data (ebcdic x'3F') that caused
an issue.
We use CCSID 37 and ascii 819.

There are more EBCDIC characters than what we see on the US Keyboard. Some
we need, such as copyright symbol, cents sign, etc, but many

We are wanting to take steps to clean the data on input, whether from ascii
or ebcdic side. We have some input already cleansed, but only at screen
program level.

Couple questions:
1. Just replacing all below ebcdic x'40'  leaves a lot of strange
characters like x'8C' (sort of a moon with a hat..). One thought is to
identify all the characters we need and replace the rest. No need to keep
line and page formatting stuff.
Is this a good idea?

2. Thinking that since a multitude of entry/update points, db triggers are
best? Am wondering about apps that write the data, and now after write, the
screen column data is different than column data in file (trigger pgm
cleaned the data - hoping to avoid opening up all the apps.

3. How far do people with heavy edi take this? Am I leaving some something
out with the keyboard characters plus a few more? These are names,
addresses, notes (which are sometimes pages of notes).

Jim Franz