client-server

linas@linas.org linas@linas.org
Fri, 22 Dec 2000 13:09:10 -0600 (CST)


It's been rumoured that David Merrill said:
> 
> learning some more from you. I work for an asp that does medical
> billing. So, we have patient records as well as clinical information
> and billing information for many different companies, often
> competitors, all in the same database.

Well, I know that e.g. with IBM DB2, you would log in with a triplet,
database-name, username, password.   The nameing scheme was such that
the databasename might be remote (some mainframe), or one of many
different databases on the local machine.  Each 'database' was in its
own subdirectory in the file system.

I know that with msql (mini-sql), you could configure it so that
the server listened to whatever portnumber, and put its files in
whatever directory.  So you could have multiple instances running
on the same machine, listening on different ports, running to
different directories, even under differnt unix uid/gid's
different config files, the whole works. 

Can't say I'd know what it takes to do this with postgres, but ther
is discussion in the postgres community about 'embedded' postgres,
that would allow multiple applications & users on the same machine to
get thier own truly private, truly unique copies.

Note, the beauty of the msql (and maybe postgres) scheme is that 
you could hand out differnt unix uid's/gid's to different db admins
and they couldn't snoop, even by accident or maliciously, on each
other's data.  Basically, you get to use the unix authentication & 
authority model, in addition to whatever the DB provides.

--linas