Subject RE: [IB-Architect] Database names
Author Jim Starkey
At 11:27 AM 5/1/00 -0400, Leyne, Sean wrote:
>The only thing I would change is the location for the alias information.
>I would suggest using isc4.gdb for two reasons.
>Keeping isc4.gdb would enable better backward compatibility for security
>functions (would still be the default security mode), until developers
>get a handle on your plug-in security API.
>Using a 'standard' db file, would provide some basic protection to the
>configuration information. Access to the alias tables should restricted

The basis of the isc4.gdb authentication model is that all applications
resident on a server have the same client base and that that client
base is different from the OS user base for the server. This model
isn't a bad compromise for many client server applications, but is
far, far from a general solution. Nobody has suggested, and I would
oppose, dropping this model from the product set, though it may be
relegated to one of several several alternatives. In any case,
it don't think it makes sense build on what may become an option.

Although there is very little enthusiasm for the Windows registry,
the registry is the OS vendor's mechanism for this type of information.
As a general policy, I think it is best to follow the host operating
system conventions and standards unless there is a compelling reason
not to do so.

The sticky question is always whether to do things the same way
on all platforms or to follow the host platform conventions. The
choice is never easy and sometimes there isn't a right answer.
One of the reasons that Interbase never implemented a Mac version
is that cross platform products either a) adhere to cross platform
consistence, in which case they get trashed in the Mac press, or
b) adhere to host platform consistence, in which case they get
trashed by corporate IS departments.

I'm perfectly willing to let the debate continue, but at the moment
I favor using the registry on windows and inventing a scheme on
Unix until the Unix world recognizes the need for a registry
equivalent (or somebody tells me that Unix has had one for five
years and I'm just to ignorant to recognize it).

Jim Starkey