Re: Informix vs. Sybase vs. Oracle vs. (gasp) MS SQL Server
From: Anthony Mandic <no_sp.am_at_agd.nsw.gov.au>
Date: 1997/11/27
Message-ID: <347D3988.153D_at_agd.nsw.gov.au>#1/1
Date: 1997/11/27
Message-ID: <347D3988.153D_at_agd.nsw.gov.au>#1/1
Snorri Bergmann wrote:
> > You can use a non-server generated lock instead. Then it
> > just becomes a matter of updating the master table. The
> > action of this update is also the action of releasing the
> > lock. No other process is locked out of reading the record
> > but others can't update or delete while that record is marked
> > as being in use (provided that they observe the rules).
>
> I have never heard before of "non-server generated locks". Could you
> explain to me what they are and how they work?
Basically its the application that generates its own lock by either setting a flag in a key table or making a row entry in a dedicated table to signal its intentions to others that it intend to alter data. The main drawback is that all others have to observe the same practice (for obvious reasons). The advantages are that no server locks are generated until the real transaction, so others can still read the data if required. If an app dies this kind of lock is retained so its useful for debugging, and if a server dies all the details would still be there (server locks would disappear). You can also record other useful information in this 'lock table' row. I.e. who, when, etc.
-am Received on Thu Nov 27 1997 - 00:00:00 CET