XPost: comp.databases.ibm-db2   
   From: news.hans@telus.net   
      
   michael newport wrote:   
      
   > correction you thought that you had made your point...   
   >   
   > You can also use JAVA on Ingres.   
      
   But nothing I've read says you can use Java IN Ingres. As in stored   
   procedures.  But since Ingres is missing triggers, I suppose it doesn't   
   really matter.  (Have to admit the rough counterpart - events - is neat.  A   
   bit like database level triggers.)   
      
   >   
   > 1. Security model                     – same   
      
   Agreed, if you restrict yourself to simple grants.  The Oracle security   
   model has a few additional things that are relevant and somewhat more   
   advanced than what the Ingres DBA manual indicates. At the tip of the   
   iceberg we see Virtual Private Database ...   
      
   > 2. Scalability                        - same   
      
   I can't seem to find any reference to scalability in the Ingres   
   documentation, or to system limitations.  Nor did Google give any   
   references to big Ingres implementations.   
      
   Pointers would be appreciated to indicate that Ingres can handle 3000   
   concurrent users and 20TByte of raw data.   
      
   > 3. Performance                        - same   
      
   Checked TCP.org - no Ingres in sight.  Any suggestions?   
      
   > 4. Shared Everything Architecture     - equivalent   
   > 5. RAC                                - equivalent   
      
   So you are saying I can have 2 servers updating the same database   
   concurrently?  Not SMP - separate machines.  Updating the same table?   
      
   Couldn't find that in the docco.   
      
   What I did find, in the System Administrator's guide (pp11-1) is "The Ingres   
   High Availability Option is not scalable; that is, it does not provide   
   active instances on multiple nodes."   
      
   > 6. DataGuard                          - equivalent   
      
   I simply could not find the terms failover and failback in the docco.   
   Pointer?   
      
   The System admin manual does discuss a cluster-coordinated switch over,   
   using scripts.  I assume you mean that.  A bit like the pre-DataGuard 'Fail   
   Safe' back in 7.3.4 and 8.0.   
      
   > 7. RMAN                               - equivalent   
      
   Nothing I saw in the docco indicated that there is a facility to do the   
   backupand track the location of the pieces of the backups to provide   
   recommendations about which  files (Journal or other) are required to   
   recover the database.  Seems it's a manual effort.  Perhaps I'm wrong?   
      
   > 28. TAF (transparent application failover) - equivalent   
      
   That's usually a function related to the cluster-cordinate failover.  TAF   
   can provide transparent failover, with no need to restart the transaction.   
      
   > 8. User defined indexes               - same   
      
   We store non-traditional datatypes (keyword, spatial data, images, music,   
   sheet music ....) and want to create a custom index?  Oracle permits that -   
   you define the indexing mechanisms and tell Oracle to use that WHILE   
   keeping the base integrity of the index mechanism.   
      
   Ingres, being open source, allows you to totally rewrite indexing - so that   
   is the same.  However, the engine doesn't then guarantee the integrity of   
   other kinds of indexes while one is putzing with the base indexing code.  A   
   bit of a trade-off?   
      
   > 9. User defined operators             - same   
   > 10. User defined locking              - nice but never needed   
      
   Most developers depend on table serialization to ensure that operations   
   block appropriately.  Even if that cuts scalability to 5-10 concurrent   
   users.   
      
   I prefer having some mechanism other than data locking to coordinate   
   concurrent operations.   
      
   > 11. Domain indexes                    - nice but never needed   
      
   Domains are basically previously undefined datatypes, somewhat like UDTs   
   (which are not limited to just "structures")   
      
   Why would anyone want to create any new data type and create an index type   
   that's relevant to the UDT?  Much better to put that code in the   
   application than in the database!  (Not)   
      
   > 12. Reverse-key indexes               - same   
      
   Didn't see that in the online SQL manual as part of the CREATE INDEX   
   command.  Have to take your word for it.   
      
   > 13. Compressed indexes                - same   
      
   Yup.  Finally one that seems similar.  With the level of flexibility   
   described, possibly even better than Oracle's compression.   
      
   > 14. Function based indexes            - nice but never needed   
      
   I think you missed the definition.  In Oracle, the index is based on a user   
   defined _expression_ - so the ability to create an index on   
      
           (col1 * col2 + col3 )   
      
   is permitted.  This are useful if the expression (using any function, even   
   user created functions) happens to occur frequently.  Why look up the   
   pieces and assemble them later?   
      
   I could go on, but ....   
      
      
      
   My conclusion now is the same as I'd concluded that early 90's when a buddy   
   went to work for Ingres and encouraged a 'fair evaluation':  Ingres is a   
   good database for reasonably plain, small, simple database usage with   
   nicely defined data buckets.  For that it's probably one of the better   
   databases, but from what I can tell all effort goes into the application to   
   overcome the limitations.   
      
   However, my philosophy is 'give unto the database everything that can be   
   centralized'.  Oracle's philosophy of additional tools, tricks and   
   simplifications help me.  They will optimize and maintain those leaving me   
   free to worry about the application.  And since I am willing to use them,   
   these items make the application's 3-year cost fairly reasonable.   
      
      
   I'm happy that Ingres is Open Source, even under CA's 'special' license. (I   
   hate it when the lawyers have to add value to a perfectly reasonable GPL.)   
      
   Of all the Open Source RDBMSs, I think Ingres will give Oracle the biggest   
   run for it's money if it survives.  However I see MySQL, PostgreSQL - and   
   now Ingres - are competing and I think that will likely devolve to the old   
   unix SysV vs BSD core battles which helped no one (except Microsoft).   
      
      
      
   Since the questions I asked were mainly rhetorical, hopefully this will end   
   the "Ingres is good too!" contribution to the DB2 vs Oracle thread.    
      
   /Hans   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)