RE: Anybody know a quick and easy way to generate a unique IFS directory for temporary use?

[jmichael.smith@xxxxxxxxxxx]

Scott

Do you have an example of the prototype for the tmpnam() ?

Michael  

-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Scott Klement
Sent: Friday, July 07, 2006 3:09 PM
To: Midrange Systems Technical Discussion
Subject: Re: Anybody know a quick and easy way to generate a unique IFS
directory for temporary use?


> First thing I looked at. As I understand the docs, it's 
> QTEMP-specific, and therefore NOT guaranteed to be unique as an IFS 
> name.

Where did you get that idea from?   There are two versions of tmpnam(), 
just as there are two versions of fopen().  One is an IFS version, the
other a native library version.

The native library version creates a unique name in QTEMP, the IFS version
creates a unique name in the /tmp directory of the root filesystem.


> I've looked into the "timestamp" idea, and I think I've got a quicker, 
> easier version thereof: just take the binary integer time() returns, 
> make a character string out of its digits, prepend some prefix like 
> "TEMP", so it's recognizable, and use that. No need to convert to a 
> human-readable time, no need to replace punctuation.

But, if two copies of your program are run at the same time (within a 
second of one another) they'll generate the same filename.  Since you 
don't have any sort of a job number included, it could be two users 
running the same menu option within a second of one another.  This could 
create some really nasty errors... you know the type, they nly pop up 
occasionally, and nobody knows why, and they never show up while testing 
in development, only in production.

PLEASE don't use the time when you need something that's "unique".  Very 
few things are less unique than the time!