Re: extending finder in gworkspace

[Rogelio M . Serrano Jr .]

Berkeley db has several access methods and a flat file is just one of 
them.

On 2003-11-18 23:25:09 +0800 Larry Cow <no@spam.org> wrote:

> On Tue, 18 Nov 2003 22:53:50 +0800
> Rogelio M.Serrano Jr. <rogelio@smsglobal.net> wrote:
>
> > I checked up locatedb. It is totally unsuitable to my needs. It uses 
> > a flat 
> > file for a database. It will not scale very well. Im thinking of 10
> > to 15 million items.
>
> Doesn't Berkeley DB use a flat file too? IMHO, the
> feature isn't valuable enough to add some more requirements to gstep.
>
> Beside, if you really want this kind of feature, you should look to
> database-inspired filesystems, like the one from BeOS.

I dont have that option. But does BeFs support queries?

I have tried using ext2. i had a humongous directory with no 
subdirectories.
the filenames were the primary keys and the links the secondary keys.
for each file i extracted all the relevant keywords and made them links
pointing to the email file. i made my own name mangling scheme like 
c++ to
handle duplicate keys. i did this for everything i put in that 
directory. mp3's,
mpegs, source code, tar archives. It worked and it still works now. I 
can
make a boolean search any way i like. It just slows down my whole 
system
when i search though and I lose some items every now and then and once
a week fsck takes a damn long time. backing up is slow too.