I figured this code would be executed repeatedly, so performance would matter.

For strings with very few surrounding spaces,

length=300
           Rate  merlyn ikegami
merlyn   8132/s      --    -80%
ikegami 41078/s    405%      --

length=500
           Rate  merlyn ikegami
merlyn   1660/s      --    -91%
ikegami 18224/s    998%      --

length=1000
          Rate  merlyn ikegami
merlyn   488/s      --    -92%
ikegami 6095/s   1149%      --
[download]

I wrote a more comprehensive benchmark, but I'm getting inconsistent results at the moment. I'll run on it a more stable machine tomorrow.

use strict;
use warnings;

use Benchmark qw( cmpthese );

my %tests = (
   ikegami11 => 's/^(.{256,}?)\s+\z/$1/s, s/^\s*(.{256}).*\z/$1/s     
+           if length > 256;',
   ikegami21 => 's/(?<=.{6})\s+\z//s,     s/^\s*(.{256}).*\z/$1/s     
+           if length > 256;',
   ikegami31 => 's/.{6}\K\s+\z//s,        s/^\s*(.{256}).*\z/$1/s     
+           if length > 256;',
   ikegami12 => 's/^(.{256,}?)\s+\z/$1/s, s/^\s+(?=.{6})//s, s/(?<=^.{
+6}).*\z//s if length > 256;',
   ikegami22 => 's/(?<=.{6})\s+\z//s,     s/^\s+(?=.{6})//s, s/(?<=^.{
+6}).*\z//s if length > 256;',
   ikegami32 => 's/.{6}\K\s+\z//s,        s/^\s+(?=.{6})//s, s/(?<=^.{
+6}).*\z//s if length > 256;',
   ikegami13 => 's/^(.{256,}?)\s+\z/$1/s, s/^\s+(?=.{6})//s, s/(?<=.{6
+}).*\z//s  if length > 256;',
   ikegami23 => 's/(?<=.{6})\s+\z//s,     s/^\s+(?=.{6})//s, s/(?<=.{6
+}).*\z//s  if length > 256;',
   ikegami33 => 's/.{6}\K\s+\z//s,        s/^\s+(?=.{6})//s, s/(?<=.{6
+}).*\z//s  if length > 256;',
   ikegami14 => 's/^(.{256,}?)\s+\z/$1/s, s/^\s+(?=.{6})//s, s/^.{6}\K
+.*\z//s    if length > 256;',
   ikegami24 => 's/(?<=.{6})\s+\z//s,     s/^\s+(?=.{6})//s, s/^.{6}\K
+.*\z//s    if length > 256;',
   ikegami34 => 's/.{6}\K\s+\z//s,        s/^\s+(?=.{6})//s, s/^.{6}\K
+.*\z//s    if length > 256;',
   merlyn    => 's/\s$// || s/^\s// || s/.$// while length > 256;',
);

$_ = 'use strict; use warnings; local $_ = our $pat;' . $_
   for values(%tests);

for my $len (256, 260, 300, 500) {
   for our $pat (
      ' ' x $len,
      (' ' x ($len/2)) . ('x' x ($len/2)),
      'x' x $len,
   ) {
      printf("length=%u, pat=\"%s...%s\"\n", $len, substr($pat, 0, 5),
+ substr($pat, -5));
      cmpthese(-1, \%tests);
      print("\n");
   }
}
[download]