imp commit=n and rbs

So what is it that makes the rbs orders of magnitude larger than the data on an import with commit=n? Freshly bounced db, one user, hpux 11, oracle 8.1.7.2, default imp buffer, 8K db buffers, single 520M table truncated in single extent (yeah, I know, I know, no LMT on this ts), 5 indices (one insane, 420M) pctfree 10%, no odd datatypes, no optimal, single rbs extended to 3.8G is not big enough.

Works just fine with commit=y, doesn't even stress the minimal (400M) rbs. Just curious what's the deal with commit=n. Even if it were screwy enough to have a table image for each index plus each index that's only 3.2G...

jg

--
@home.com is bogus.
http://www.signonsandiego.com/news/uniontrib/thu/business/news_1b24csfb.html