RE: RE: Auth lists and object authority

midrange-l-request@xxxxxxxxxxxx wrote:

>   4. RE:  Auth lists and object authority (ISNM)
>
>  Tom said:
>   
>  "At PowerTech, we've put a feature called 'Memorized Transactions' into the 
> exit-point solution we sell -- NetworkSecurity. This provides granularity 
> down to the level of specific transactions. If the only authorized 
> transaction for a user is "SELECT FLDA, FLDB FROM FILEA", then "SELECT FLDA, 
> FLB, FLDC FROM FILEA" will be rejected. "
>   
>  So, what happens if the user submits a transaction that says "SELECT FLDB, 
> FLDA FROM FILEA" (change of the select order)?  Is the default of the product 
> such that it rejects if there is no explicit allow?  Or do you have to 
> specifically deny it for that particular file/field combination?


Steve:

That was an example of a way to have an exit program approach blanket 
restrictions to allow exceptions. In the scenario you describe, a memorized 
transaction probably isn't appropriate.

Memorized transactions (MTs) aren't intended for general ad hoc queries but 
rather for high-performance response to applications where common transactions 
show up regularly. We don't want to do syntax checking, object checking, etc., 
just so a Windows server app or web server can grab results of a query that 
might run over and over again.

Note that MTs can be generic. The transaction might be "SELECT FLDA, FLDB FROM 
ORDERS WHERE CUST = %". Trivial example, but it'll do. An app could send the 
same base query thousands of times, each for a different CUST.

If the transaction comes through as "SELECT FLDB, FLDA..." instead, then it 
should be obvious that it's no longer the _same_ app submitting the query. 
Programs send typos far less often than people. And if the action for the MT 
says REJECT, then it most definitely is rejected. Suddenly it's not the 
programmed app but something/someone impersonating the app and essentially 
sending a malformed query.

We've looked at so many ways of scanning transactions for various object 
references and keep finding holes in every technique. The amount of processing 
necessary can kill some of the servers we have to handle. But a MT can zip 
through.

For ad hoc queries, the overhead isn't as important; so MTs wouldn't be used as 
often. A customer would choose to use them only where they made sense.

We feel pretty sure that we're not going to improve on OS/400's existing 
object-level authority checking; and we're almost as sure that we're not going 
to improve the speed of parsing/analyzing SQL or FTP transactions or whatever 
type of transaction it is. Besides, OS/400 authority checking and transaction 
parsing will happen anyway; so why should we more than double the time it takes 
for every transaction, especially transactions that _need_ high speed?

So, yes, such a transaction will be rejected. I imagine that our customers who 
use them would be pretty disappointed if it weren't, eh?

Tom Liotta