I went ahead and implemented an HTML parser to produce a "DOM" in accordance with the aforementioned structure.

Here is the program to produce HTML from that "DOM":

sub htmlize{
    my ($tag_name,$dom)=@_;
    return $dom unless ref $dom;
    my $html;
    my %cnt;
    my $no_text;
    $html .="<$tag_name";
#    print "<$tag_name";
    for my $rank (0..$dom->{'_meta_attr_n'}-1){
        my $attr_name=$dom->{'_meta_order'}->[$rank];
        if($attr_name eq '/'){
            $no_text=1;
            next;
        }
        my $attr=$dom->{$attr_name}->[$cnt{$attr_name}||'0'];
        $attr = join ', ',@$attr if ref $attr;
        $html .= " $attr_name=\"$attr\"";
#       print " $attr_name=\"$attr\"";
        $cnt{$attr_name}++; # in the unlikely event that a child tag a
+nd this attr have the same name
    }
    $html .= '/' if $no_text;
#    print '/' if $no_text;
    $html .= '>';
#    print '>';
    for my $rank ($dom->{'_meta_attr_n'}..$#{$dom->{'_meta_order'}}){
        my $tag_name=$dom->{'_meta_order'}->[$rank];
        $html .= htmlize($tag_name,$dom->{$tag_name}->[$cnt{$tag_name}
+||'0']);
        $cnt{$tag_name}++;
    }
    $html.="</$tag_name>" if !$no_text;
#    print "</$tag_name>" if !$no_text;
    return $html;
}
[download]

Is this unwieldy compared to the [{'key'=>$key_name, 'value'=>$value},...] structure?