Re: ROWID Question - functionality

In article <31975126.52B3@segel.com>, Mike Segel <mikey@segel.com> wrote: > Why not use a serial field? > What you could do is create a small table which has the table id and a > serial field as > a counter. But in general, what your asking doesn.'t make sense. What is > the functionality you > are trying to achieve? > > -Mikey The functionality that I am trying to achieve is to provide notification to clients of a 3-tier client-server application of various database table updates from (2) potential sources: 1) An application server process, which I control. 2) Updates from another instances on another nodes, through a replication mechanism. (Informix CDR or Praxis OmniReplicator). The Praxis product does not support serial field types, so they cannot be used. Informix CDR does not fire triggers, so they cannot be used. My design was to write to a log table, containing the tablename (actually a enumeration code), the update type, and key fields. This is done as part of the update transaction to be monitored. The log table is replicated along with the other tables of interest. The monitoring process scans the log table, looking for new records, which is what I would like to use the rowid for. Once it reads the keys, it is able to read the actual data (for non-deletes) and notify clients. Another thought was to use a time based key, but there might be substantial delays in propagating replicated data, which I want to see coming in. It sounds like I should use rowids, but plan on dropping and rebuilding the log table as part of maintenance activities, to ensure that they (the new entries) are kept in sequence. Any other thoughts ? -- Peter Sylvester MITRE Corp. Bedford, MA (peters@mitre.org)