On Mon, Nov 12, 2001 at 01:10:13PM +0100, Oskar Sandberg wrote:
> On Sun, Nov 11, 2001 at 07:02:23PM -0500, Tavin Cole wrote:
> <> 
> > What makes the datastore so nasty to fix is that it's not a matter of
> > weeding the bugs out of a sufficient design..  there are just big
> > architectural holes that still need to be filled.  I'm not trying
> > to discourage anyone from finding and fixing bugs in the current
> > system, but I'm totally convinced that any serious effort to make
> > it really robust will lead the designer down the same paths of
> > extensive re-designing that I've been down.
> 
> Tavin, how much of the necessary stuff (circular buffers etc) could be
> implemented on top of regular files instead of the native FS? I know the
> advantages of the FS, but it has to be weighted against the workload
> necessary, and I'm beginning to feel that what started as a joke is
> growing ad inifinitum.
> 
We may be going about this the wrong way.  We should have designed a
datastore for moderate size (100-1g).  People that want to run
ridicustores should be using an SQL database or some tech thats designed
to handle this sort of situation.  We can still get the sort of security
benefits we want even using such a system, and we dont waste time writing
what is basically a high performance filesystem.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 232 bytes
Desc: not available
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20011112/0aaccfb1/attachment.pgp>

Reply via email to