auto-backup across the net

[Austin Schutz]

> Provided that the "database" is not stored directly on the webserver which
> is running the cgi, the connection state of the client/cgi is independant of
> the connection state of the cgi/database.  A client will not lose its cgi
> session if the cgi is unable to access the database.  A robust cgi can
> handle redirecting where it will make its database connection, there is
> never a need to inform the client.
> 

	No, this is not necessarily true. There may be times where the
cgi is not able to make the database connection and would want to inform the
client. Yes, there are probably ways to architect around this situation most
of the time.
	It's beside the point though. The question is not "why would you
do it?", the question is "how do you do it?".
	The question is not entirely trivial. The background process is
not always a database, where (network connectivity allowing) you can simply
point to a redundant server.  An example:

	A webserver has an attached webcam, which is faulty. A backup
webcam is hosted on a different server, administered by someone else. Is
it possible to immediately redirect the person to the new site without
user intervention?

	Austin
TIMTOWTDI