Let the data speak for itself.
Code
vinko@parrot:~$ more rand.php
<?php
$randrandsum = 0;
$randconstsum = 0;
$count = 20;
for ($j = 0; $j < $count; $j++) {
$randrand = 0;
$randconst = 0;
for ($i = 0; $i < 10000000; $i++ ){
$a = rand(1,1000);
$b = rand(1,1000);
if ($a == $b) $randrand++;
}
for ($i = 0; $i < 10000000; $i++ ){
$a = rand(1,1000);
$c = 1000;
if ($c == $a) $randconst++;
}
$randrandsum += $randrand;
$randconstsum += $randconst;
print ($j+1)." RAND-RAND: $randrand RAND-CONST: $randconst\n";
}
print "AVG RAND-RAND: ".($randrandsum/$count);
print " AVG RAND-CONST: ".($randconstsum/$count)."\n";
?>
Test run
vinko@parrot:~$ php rand.php
1 RAND-RAND: 10043 RAND-CONST: 10018
2 RAND-RAND: 9940 RAND-CONST: 10132
3 RAND-RAND: 9879 RAND-CONST: 10042
4 RAND-RAND: 9878 RAND-CONST: 9965
5 RAND-RAND: 10226 RAND-CONST: 9867
6 RAND-RAND: 9866 RAND-CONST: 9992
7 RAND-RAND: 10069 RAND-CONST: 9953
8 RAND-RAND: 9967 RAND-CONST: 9862
9 RAND-RAND: 10009 RAND-CONST: 10060
10 RAND-RAND: 9809 RAND-CONST: 9985
11 RAND-RAND: 9939 RAND-CONST: 10057
12 RAND-RAND: 9945 RAND-CONST: 10013
13 RAND-RAND: 10090 RAND-CONST: 9936
14 RAND-RAND: 10000 RAND-CONST: 9867
15 RAND-RAND: 10055 RAND-CONST: 10088
16 RAND-RAND: 10129 RAND-CONST: 9875
17 RAND-RAND: 9846 RAND-CONST: 10056
18 RAND-RAND: 9961 RAND-CONST: 9930
19 RAND-RAND: 10063 RAND-CONST: 10001
20 RAND-RAND: 10047 RAND-CONST: 10037
AVG RAND-RAND: 9988.05 AVG RAND-CONST: 9986.8
Given the above results I'd say that for all practical purposes, both options are equivalent, giving the expected 1/1000 result on both cases.