Even shorter :-) (Update: Sorry, not quite the same, see replies.)

$ perl -lp000e1  input.txt >output.txt
[download]

$ perl -MO=Deparse -lp000e1

BEGIN { $/ = ""; $\ = "\n"; }
LINE: while (defined($_ = readline ARGV)) {
    chomp $_;
    '???';
}
continue {
    die "-p destination: $!\n" unless print $_;
}
[download]

use warnings;
use strict;

local ($/,$\) = ("","\n");
while (<>) {
    chomp;
    print;
}
[download]

Shorter, yes, but it looks like you've re-introduced the "input-file is small enough to be read into memory" constraint. Or have I missed something?

Shorter, yes, but it looks like you've re-introduced the "input-file is small enough to be read into memory" constraint. Or have I missed something?

Hm, good point, I guess in my excitement over the nice short oneliner I sort of did. -000* turns on paragraph mode, so if the file contains lots of short "paragraphs" with blank lines in between, it'll be fine, but if the file contains, say, a block of several MB of non-blank lines, a single blank line, and another huge paragraph, then yes, you're right, my solution isn't that great.

* perlrun is a little unclear whether it's -00 or -000, both seem to work.