One thing to be careful of (that bit me when I did this) is that some DBMs limit the length of string that they can store. I believe I ran into a 1024-character limit, which is clearly not all that long.

You can certainly work around this; I ended up using two tables. The actual data is split up into 1000-character chunks and a separate table is used to index into that for reconstruction. A better solution would be DBM::Deep, which handles all of this for you. :-)