samba - Jul 1999 - Why does RH's Samba build use a file for shared memory?

Samba needs a block of shared memory for storing file handles and
related info.  The amount of memory to use is defined as a specific
size (default: 1MB), created at smbd startup, and used thereafter.

I've noticed that in the RPMs available from the Samba team's Website
that they use RAM to hold the shared memory (this is the default),
while RedHat's own RPMs create a file to hold that data.

I assume that RAM-based access to the Samba data is faster than 
file-based, so why does RedHat not use the default configuration of 
storage in RAM?

Thank you.


*** Steve Snyder ***

> I've noticed that in the RPMs available from the Samba team's
Website
> that they use RAM to hold the shared memory (this is the default),
> while RedHat's own RPMs create a file to hold that data.
nope, Samba 2.0.4 used a file because with Linux 2.0.XX mmap/fcntl is
faster than shmem/semops. In Linux 2.2 the SysV shmem/semops code was
greatly improved and is now much faster than mmap/fcntl so I decided
to remove the special case for Linux and now Linux uses SysV if
available and mmap if ont, just like all the other Unixes we support.

So the difference is 2.0.4 -> 2.0.5 not samba.org -> redhat.com

Cheers, Tridge