From rutgers:
    skip over label junk at front of tape
    verify recs in sequence and not duped
    interpret SCD field when printing saveset name

==> retrieve archive info!
probably doesn't deal with holes in files
't' should display last writer string
fix bogus atimes (1994)
map writer name -> uid if possible and root.
    have auxiliary table?
fix argument parsing so can combine options.
(need to skip over 2560 length (tape labels?) at front of first tape.)
(files spanning tape bounds - partially done)
(saveset index)
(select by saveset#/filename)

DONE:
==> map tops-20 filenames to unixnames better, including directories:
on 't' or v output, number the savesets and the files.
==> select by archive info: saveset#/file#
embedded null bytes
match multiple patterns -crum
match via regexp -crum
skipped user info records should only be on verbose and say directory recs
kludge the protections so remove x bits
set utimes 

CR stripping affects filelength, can leave getstring part way thru word
    due to maxbytes exceeded. - fixed.
preserve protections: use pcc20 mappings

CR stripping - partial
read from stdin
default rmt0 -> rmt8
minor speedups
create filenames in lower case
match in lower case
use perror in punt()
do directory listing on -t flag
do extract on -x flag
limit directory listing to matching files.
check for write errors

------------------------------------------------------------
fdb scanner outline:
Can all info be saved in slightly modified tar format?
Would be nice...
Here's the extra info needed:
1 bit - archive
1 bit - offline
1 bit - invisible
6 bits - bytesize
short - generation number
39 char - last writer 	    ===> Map to uid
-----
43 chars

6 chars - archive tape #1
6 chars - archive tape #2
2 shorts - #1 saveset and file number
2 shorts - #2 saveset and file number
----
20 ==> 63

[2 shorts - #writes, #refs]
[39 char - author name]
----
43 ==> 106

