| T.R | Title | User | Personal Name
 | Date | Lines | 
|---|
| 185.1 | apples and oranges | SAGE::OISPG | Keith A: DESINE at ground zero | Tue Aug 30 1988 16:02 | 20 | 
|  |     I dont know DECIntact but doubtless an enthusiastic advocate will
    jump in and tell you about it.
    
    A point on relativity: products like ACMS, DECIntact are intended to be
    TP monitors that layer over databases, operating systems, etc. Hence we
    have to consider the definition of a "transaction" at each layer to see
    what that layer gives us in terms of protection. For example, an ACMS
    user may consider a certain activity to be one logical transaction but
    Rdb may be given several Rdb transactions to do for that single logical
    transaction. Rdb will look after each Rdb transaction but knows nothing
    about the ACMS logical transaction. If Rdb transaction R1 may be
    committed and then R2 will fail and be rolled back. This may cause
    a failure for the ACMS logical transaction L1 but there is no way
    (now) that Rdb nows that the previously committed R1 is invalid.
    
    This difference in meaning between transactions is the reason for
    discussion on 2 phase commit. Customers want a system where a high
    level component (like ACMS, DECintact) can easily coordinate
    transaction commits and rollbacks with lower levels like Rdb, RMS,
    DBMS etc.. 
 | 
| 185.2 | Logical Transaction | CSTEAM::WADSWORTH | KIRBY WADSWORTH | Tue Aug 30 1988 22:06 | 6 | 
|  |     Thanks .1
    As an old OLTPer, I view a transaction as a logical unit of work, not
    an  individual interaction with a database.  If I transfer money
    from my checking to savings, I want to be assured that the money
    went into my savings account after it was successfully removed from
    my checking account.   
 | 
| 185.3 | I don't think that's why we have DecIntact | WIBBIN::NOYCE | Bill Noyce, Parallel Processing A/D | Thu Sep 01 1988 15:02 | 10 | 
|  |     DEC database products have provided transaction protection for a
    long time: VAX DBMS V1 had it in 1981; VAX Rdb/VMS V1 in 1984.
    RMS didn't have it until RMS Journaling (1987?).  Intact was developed
    to fill this hole in RMS (and also to add some other things (hashed
    access?) that you probably know more about than I do).  DEC decided
    to buy Intact and re-sell it for reasons I don't entirely understand,
    but I'm sure that it's partly because it appeared to provide much
    better performance than ACMS did.  I don't think it gives any more
    transaction protection than are available in the database products
    and (finally) now RMS.
 | 
| 185.4 |  | CSC32::STENOISH | Jim Stenoish, VIA TBU, CSC/CS | Thu Sep 01 1988 19:23 | 7 | 
|  |     DECintact allows transaction integrity over multiple RMS databases,
    something neither Rdb or DBMS currently allow.  At the time DECintact
    was acquired, ACMS had no transaction integrity if your database
    wasn't all in 1 file Rdb or DBMS database.  Of course, DECintacts
    rollback/rollforward recovery facilities is not currently integrated
    with Rdb's or DBMS' recovery facilities, so there doesn't seem to
    be any advantage to using DECintact with Rdb or DBMS.
 |