> On Wed, Feb 06, 2002 at 11:48:17AM -0500, Ian Clarke wrote:
> > Thumbs down.
Ok. I will back out my changes tonight.  I do so under protest.  But a vote 
is a vote and I have clearly lost.

> >
> > People shouldn't need to test insertions in this manner, if an insertion
> > didn't occur correctly, then fcp should return an error.  This command
> > encourages behavior that should not be nescessary.

Boats should not leak. Most boats leak. If you pump them the are perfectly 
usable.  If you stop pumping them most eventually sink. Just because your 
boat shouldn't leak, does that mean you should let it sink?

So, even if your facts were correct, I wouldn't agree with you.  But they are 
not.

In order for fcp to "know" that the data ended up somewhere in the network 
where it is reasonably retrievable, it would have to have global knowledge of 
the network, which it doesn't have.  The only way to be sure something is 
retrievable from the network is to try to retrieve it.  The current 
implementation can't do that if the key is in the data store. 

As Tavin has already mentioned, another way to solve this problem is to put 
an option on ClientGet and ClientPut to delete the key locally before the 
request/insert.  That seems like a politically more viable solution to the 
problem that I was trying to solve.

--gj

-- 
Freesites
(0.3) freenet:MSK at SSK@enI8YFo3gj8UVh-Au0HpKMftf6QQAgE/homepage//
(0.4) freenet:SSK at npfV5XQijFkF6sXZvuO0o~kG4wEPAgM/homepage//

_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to