It sounds like you've been drinking Microsoft's kool-aid -- don't let the user 
do what they need to's too dangerous.  Just try and print a directory 
list from your file browser in Windows for example (not that it's 
dangerous..but it's pretty obvious someone might want to do it).
Software should ALWAYS allow for an expert mode.  Where we can do dangerous 
things when we need to (for example keeping your journal on local file storage 
when your database is on network storage).
Just like it's stated that it's dangerous to use network storage but you allow 
it nonetheless.  That is MUCH more likely to happen that someone trying to 
store the journal elsewhere.  So how do you justity allowing that to occur?  
Seems you're being a bit non-orthogonal.
If you really needed to recover such a database you just have to move the 
journal file to the database directory.  
Michael D. Black
Senior Scientist
Northrop Grumman Mission Systems


From: on behalf of Richard Hipp
Sent: Wed 7/14/2010 12:51 PM
To: General Discussion of SQLite Database
Subject: EXTERNAL:Re: [sqlite] Sqlite Insert Speed Optimization

On Wed, Jul 14, 2010 at 1:34 PM, Black, Michael (IS) <
> wrote:

> Was that a facetious remark???
> Rather than "here's a function/pragma that allows you to put the journal
> file where you want -- but BE CAREFUL BECAUSE..."
> Writing you own VFS is not for the casual user...
> I was trying just to find where the journal filename was created but
> there's no "db-journal" string in either the .h or .c file.
> It would be trivial to add a function to set it via the C interface.

Putting the rollback journal in any directory other than the same directory
as the database file is an dangerous thing to do, because it risks being
unable to locate the rollback journal after a crash, resulting in database
corruption.  Hence, we have no intention of supporting such a feature.  If
you really need it badly enough, you can write your own VFS to make it
happen.  Yes, writing your own VFS is hard to do.  But this serves to
discourage people from doing it, which is what we want.

D. Richard Hipp
sqlite-users mailing list

sqlite-users mailing list

Reply via email to