CountZero's solution is of course far more elegant. I thought I'd see if I could code a filter using just regex, however:

use strict;
use warnings;

my ($xml, $content, @xml);

$xml = join '', <DATA>;

while ($xml =~ m/(<details>.*?<\/details>)/sg) {
    $content = $1;
    next if $content !~ /<address>.+?<\/address>/;
    push @xml, $content;
}

$xml = "<specification>\n" .
join("\n", map { "  $_" } @xml) .
"\n</specification>";

print $xml;

__DATA__
<specification>
  <details>
    <name>johan</name>
    <address>Langgt 23</address>
    ---more info---
  </details>
  <details>
    <name>venu</name>
    <address>storgatan 27</address>
    ---more info---
  </details>
  <details>
    <name>kent</name>
    <address>nygatan 46</address>
    ---more info----
  </details>
  <details>
    <name>johan</name>
    <branch>ece</branch>
    ---more info--
  </details>
</specification>
[download]