I'll show you another common way that this parsing problem is handled. May be useful in other situations...

When you see the "start-of-record", call a subroutine to process the record. In this case there is no state variable to say that "we are in the record", the fact that you are in the process_record() subroutine serves that purpose.

#!/usr/bin/perl -w
use strict;

while (<DATA>)
{
    process_record() if /^START/;
}

sub process_record
{
   my $line;
   while ($line = <DATA>, $line !~ /^END/)
   {
      print "$line"
   }
   print "\n";  #a printout spacer for next record
}

=prints
These are the first
set of lines
which are to be extracted

These are the second
set of lines
which are to be extracted

=cut

__DATA__

XXXX
YYYY
START
These are the first
set of lines
which are to be extracted
END
XXX
ZZZ
YYY
START
These are the second
set of lines
which are to be extracted
END
aasds
tteret
tertetr
[download]