Edit report at https://bugs.php.net/bug.php?id=47890&edit=1

 ID:                 47890
 Updated by:         yohg...@php.net
 Reported by:        bernard dot fouche at kuantic dot com
 Summary:            Function 'uniqid()' has a design flaw
-Status:             Open
+Status:             Verified
 Type:               Feature/Change Request
-Package:            Feature/Change Request
+Package:            *General Issues
 Operating System:   Any
-PHP Version:        5.2.9
+PHP Version:        master
 Block user comment: N
 Private report:     N

 New Comment:

It might be good to set more_entropy set to true for next PHP. W/o combined 
lcg, 
it has too weak uniqueness.


Previous Comments:
------------------------------------------------------------------------
[2009-04-03 16:20:50] bernard dot fouche at kuantic dot com

Description:
------------
IMHO 'uniqid()' does not provide any warranty about the returned id being 
really 'uniq':

- uniqid() just calls 'gettimeofday()' (from php-5.2.9/ext/standard/uniqid.c): 
there is no warranty that the underlying operating system is able to really 
count microseconds. May be it counts by chunks of 10us, or 1ms, etc..

- With multi-core CPU and since AFAIK gettimeofday() is not a 'real' system 
call any more (on Linux) but reads from shared memory updated by the kernel, 
chances to have more than one process or thread being able to get the same 
value are much higher than with older hardware and OS implementations.

- NTP may update the system time and whack the microsecond counter at any time.

I think that uniqid() should rely on other information, like the process ID, 
thread ID, and an internal counter... Sure the programmer can use the 'prefix', 
but then what is the purpose of this function if the programmer has to take 
care about these details?

Reproduce code:
---------------
---
>From manual page: function.uniqid
---




------------------------------------------------------------------------



-- 
Edit this bug report at https://bugs.php.net/bug.php?id=47890&edit=1

Reply via email to