Jim Starkey wrote:

>
> At 01:23 AM 4/29/00 -0400, Claudio Valderrama C. wrote:
> >
> >Why this couldn't be achieved creating a new table inside isc4.gdb? Do you
> >want to see this global db removed in the future and replaced totally by a
> >per-db security table(s)?
> >
>
> Yes. I would like to see isc4.gdb removed from the engine (where it
> breaks the layering)

[...]
Yeah!

> The "right" place for the mapping between database names and
> database filenames is the registry for Windows systems and a
> file in either the Interbase install directory or possibly
> /etc on Unix (suggestions solicited).
>

Interbase install dir. There's already plenty of stuff in /etc, and
we already have to know where the interbase directory is.

> The logical translation point is the Y-valve (why.c). Among
> the algorithms that could be used:
>
> 1. Assume a connect string is a name if it doesn't contain
> a dot.

Or perhaps if it doesn't contain any directory separator characters.
I never thought that connecting to host:mydatabase.gdb was that
great a feature.

>
> 2. Try to translate the connect string, regardless. If it
> doesn't match a name, assume it's a file.
>

This seems cleaner than option 1. It might be good to not allow
database names to have / or \ in them, so you couldn't have
a database >alias< \foo\bar\my.gdb and a >database< c:\foo\bar\my.gdb.

> My druthers would be to make database names case insensitive to
> match SQL semantics. Thoughts?

Seems fair.
[...]

--
Reed Mideke
rfm(at)collectivecomputing.com