At 08:03 AM 3/29/00 -0800, you wrote:

>From: "Markus Kemper" <mkemper@...>
>
>
> Currently InterBase 'Shadow' technology is bound to
> the node executing the InterBase engine. Would the
> ability to 'Shadow' to a remote node:
>
> a) be difficult to implement? (my guess is no)
> b) greatly enhance our failover story?
>

Why is it are shadows bound to the server node? Is this
a feature that could be implemented by removing code?

The requirement that the primary file be on the server
node was required to ensure that two servers don't try
to use the same file under NFS, and was not present on
either VMS or Apollo. In the case of SuperServer which
can take an OS exclusive lock on the file, it is completely
unnecessary.

The other thing that would help failover would be a registry,
preferrable replicated, where a client could go to find
the current server. Perhaps we should consider a database
name (syntactially differentiable from a file name) to
somebody could transparently map to s (server, filename)
pair.

Jim Starkey