Server Unique Identifier Question

[David T. Ashley]

Hi Guys,

In my application I need a way to generate an identifier that is guaranteed
to never recur in the lifetime of the server. The method I'm using now is
to concatentate:

a)Integer time, in seconds,

b)Micro time, in microseconds, and

c)PID

into one string, then to spin-lock or micro_sleep() until the micro_time
changes.

My assumption is that this identifier will be unique because no two
processes can have the same PID at the same time.

Is there a better way?

Thanks, Dave.

P.S.--Application: Linux, Apache, PHP, MySQL, web database stuff.

[David Haynes]

Re: Server Unique Identifier Question

David T. Ashley wrote:[color=blue]
> Hi Guys,
>
> In my application I need a way to generate an identifier that is guaranteed
> to never recur in the lifetime of the server. The method I'm using now is
> to concatentate:
>
> a)Integer time, in seconds,
>
> b)Micro time, in microseconds, and
>
> c)PID
>
> into one string, then to spin-lock or micro_sleep() until the micro_time
> changes.
>
> My assumption is that this identifier will be unique because no two
> processes can have the same PID at the same time.
>
> Is there a better way?
>
> Thanks, Dave.
>
> P.S.--Application: Linux, Apache, PHP, MySQL, web database stuff.
>
>
>[/color]
Do a google search for uuid to find the proposal by Microsoft. There are
many Linux implementations of this function.

-david-

[Sean Barton]

Re: Server Unique Identifier Question

maybe doing all that is a little overkill. why not use mt_rand? and if
you must hash the result. if you use a large enough number for the
random function you are pretty much looking at a 0 chance of it
recurring

hope this helps
Sean Barton

[noone]

Re: Server Unique Identifier Question

David T. Ashley wrote:[color=blue]
> Hi Guys,
>
> In my application I need a way to generate an identifier that is guaranteed
> to never recur in the lifetime of the server. The method I'm using now is
> to concatentate:
>
> a)Integer time, in seconds,
>
> b)Micro time, in microseconds, and
>
> c)PID
>
> into one string, then to spin-lock or micro_sleep() until the micro_time
> changes.
>
> My assumption is that this identifier will be unique because no two
> processes can have the same PID at the same time.
>
> Is there a better way?
>
> Thanks, Dave.
>
> P.S.--Application: Linux, Apache, PHP, MySQL, web database stuff.
>
>
>[/color]


Just curious, but why not use $_SERVER['UNIQUE_ID']?

M.

[Colin McKinnon]

Re: Server Unique Identifier Question

David T. Ashley wrote:
[color=blue]
> to never recur in the lifetime of the server. The method I'm using now is
> to concatentate:
>
> a)Integer time, in seconds,
>
> b)Micro time, in microseconds, and
>
> c)PID
>
> into one string, then to spin-lock or micro_sleep() until the micro_time
> changes.[/color]

There's not a *better* way, although your proposal is not very efficent (80+
bits a time / 18 decimal digits). (I would have used a counter with a
mutex)

C.

[David T. Ashley]

Re: Server Unique Identifier Question

"noone" <noone@nowhere. com> wrote in message
news:sErFf.2234 4$Jd.12947@news svr25.news.prod igy.net...[color=blue]
>
> Just curious, but why not use $_SERVER['UNIQUE_ID']?[/color]

Thanks for that tidbit. I wasn't aware that this was available.

From

http://httpd.apache.org/docs/1.3/mod/mod_unique_id.html

I glean that the approach I'm using and the approach Apache is using is
identical except I use microtime and a spin lock whereas Apache uses a
counter.

So, it reinforces my belief that my approach is correct.

I wasn't aware that Apache does this.

Thanks!

[David T. Ashley]

Re: Server Unique Identifier Question

"Colin McKinnon"
<colin.thisisno tmysurname@ntlw orld.deletemeun lessURaBot.com> wrote in
message news:uhuFf.6096 $Dn4.2149@newsf e3-gui.ntli.net...[color=blue]
>
> There's not a *better* way, although your proposal is not very efficent
> (80+
> bits a time / 18 decimal digits). (I would have used a counter with a
> mutex)[/color]

OK, let's say I do "counter with a mutex".

If I use MySQL to hold the counter, the mutex mechanism is obvious ("LOCK
TABLE ...").

However, if I hold the counter in a file, what is the best mutex mechanism
in PHP?

Thanks, Dave.

[David Haynes]

Re: Server Unique Identifier Question

David T. Ashley wrote:[color=blue]
> "Colin McKinnon"
> <colin.thisisno tmysurname@ntlw orld.deletemeun lessURaBot.com> wrote in
> message news:uhuFf.6096 $Dn4.2149@newsf e3-gui.ntli.net...[color=green]
>> There's not a *better* way, although your proposal is not very efficent
>> (80+
>> bits a time / 18 decimal digits). (I would have used a counter with a
>> mutex)[/color]
>
> OK, let's say I do "counter with a mutex".
>
> If I use MySQL to hold the counter, the mutex mechanism is obvious ("LOCK
> TABLE ...").
>
> However, if I hold the counter in a file, what is the best mutex mechanism
> in PHP?
>
> Thanks, Dave.
>
>
>[/color]
What's wrong with flock()?