For instance, with this:

print for
    map { substr $_, 30 }
    sort
    map { sprintf "%-30s%s", (split / /, (split /\|/)[1])[1], $_ }
    <DATA>;

__DATA__
123|Pete Smith|4321|321G4
2134|Mike Malarky|93F821
83|Dave Waylaid|8374W
[download]

Output:

2134|Mike Malarky|93F821
123|Pete Smith|4321|321G4
83|Dave Waylaid|8374W
[download]

The difficult part is that nested split. The inner one, (split /\|/)[1] selects the second field in your data. Then the result is fed to the outer, (split / /, inner)[1], which extracts the second word. That is mapped to a string like this: Malarky                       2134|Mike Malarky|93F821. This string is ASCIIbetically sorted and then, the right part of it is extracted with substr and returned to print. This technique (mapping to a string, sorting and extracting the previously mangled data) is called Guttman-Rosler transform, google for it.

As a side note, I'd like to point out that the first line in your example data appears to have four fields separated with a pipe, while the rest have only three fields.