Re: [Firebird-devel] Boot kits & gpre

[John B. <bel...@cs...>]

On Friday, May 18, 2001, at 11:10 AM, Ann W. Harrison wrote:

> At 10:40 AM 5/18/2001 -0700, John Bellardo wrote
>
> <met.e/met.c
>
>> It does more that just use fields.h and relations.h.  It can extract
>> the metadata for user relations, which requires db access.
>
> Right it extracts whatever metadata is required.  The engine, gbak &
> gfix require only the system table metadata.

Since met.e is part of the engine (and the only .e file gpre uses) then 
gpre should [also] only require the system table metadata.

>
>> It seems to me (I could be wrong here) that met.e has to functions.  
>> One
>> is to access the meatdata, and to other is to access the meta-metadata.
>
> Err... sort of.  It does have a built-in knowledge of RDB$RELATIONS,
> RDB$FIELDS, etc., so it's not rebuilding itself from first principles.

Thats what I was trying to get at.  The fact that it already has enough 
knowledge built in to compile the engine without having an engine 
installed.

>
>
>> One solution is to redesign the entire security model, disposing of 
>> isc4.gdb :)
>
> Sounds good.

It's on the list, but that list is growing faster than it is shrinking :(

>
>> More realistically (at least in the short term) though I was under the
>> impression that Windows only run SS, which might be the problem.  In
>> *nix you can run gbak against Classic (which is how the Darwin build 
>> works).
>> The Classic engine will assume your SQL identity is your login, and, 
>> using
>> your permissions create an isc4.gdb file.
>
> The problem isn't classic vs SS.  It's the uSoft security model.  On
> Unix, you login, you have an identity, and the database can verify your
> identity.  On Windows, you pick up the computer and start typing - no
> login.  No identity.  Of course NT is different, but...

I've always felt that Windows (not NT) had the security model of 
"everyone's root".  So IMHO if there is no isc4.gdb file the Windows 
build (not NT) should follow that model.

-John