Brad,

> > 2. Ever, ever use the setting FORCED WRITES ON!!!
>
> Do I understand this properly in that the default on Linux is
> to *not* have
> forced writes on but that it is recommended to turn it on?

You are correct, the default for all non Win32 platforms is FORCED
WRITES = OFF.

> If so, is there
> any way to turn it on when the database is created or using
> the API rather than running "gfix -write sync"?

The DML syntax doesn't support the ability to specify the Forced Write
state on create.

As for the API, I don't know -- I'll leave that for someone else.

> Also if its recommended on Linux, why is it not the default?

I guess it comes down to a question of overall safety.

The cases of a Win server abending for no particular reason are _well
known_, accordingly it is definitively NOT safe to run without FORCED
WRITES = OFF (expect for the sado-masochistic).

The same is not at all true for Unix/Linux (at least so I have been
told, these many years of following the IB lists), the chances of a
server abending are less than 1%. Accordingly, running with FORCED
WRITES = ON would 'impose' too high a performance penalty, and thus it
was decided many moons ago to not default the setting to ON.

Personally, I would **never** run a DB server without FORCED WRITES =
ON, regardless of an performance penalty -- the risk of data loss is not
worth it. With today's cached SCSI and IDE disk controllers (battery
backup option is MANDATORY) there would be absolutlely no performance
penalty.


Sean