> On Sun, 28 Nov 2004 21:21:07 +1100, Helen Borrie <helebor@t...>
> wrote:
>
> > At 11:49 AM 28/11/2004 +0200, you wrote:
> >
> > It's assumed that it always does detect it - monitoring
> > attachments is part of what the server does. It will continue
> > listening for a response from a connection for a set timeout
> > period (the length of which depends on the NOS
> > configuration),

I think, it will be good if this assumption is clearly stated in the
docs.

> > after which it assumes the connection is dead
> > and proceeds to deal with unfinished transactions from that
> > connection - rolling them back or flagging them as limbo,
> > according to whether the transactions were
> > related to just that connection or involved other connections,
> > respectively.
> >

For the limbo case, why does the server have to flag it as limbo? Is
the server not being able to just rolling back it? If the server can
not just roll back limbo transactions, how should we clear these
limbos? Will gfix -sweep take care of it?

>
> Thanks for explanation :) , just last small question: what is 'NOS'?
>

I also thank Helen and also Alan for the thorough explanation.

--
Bambang