Re: Callable statement with RPG

["Larry Loen" <lwloen@xxxxxxxxxx>]

>>
>> Wouldn't one or two nasty bugs be a sufficient teacher?
>>
>One or two nasty bugs ... did you ever have problems with deadlocks in
>multithreaded applications???

Maybe that's my problem.  I've been a developer of system programming for
nearly 30 years, including the original single level store paging code.
Needless to say, I have significant experience with such bugs both at the
system and application level.

Perhaps your suggestions are practical for your environment, but it seems
very draconian to me.  Java's synchronization is itself pretty draconian
(and therefore simplified compared to what I deal with), which is why your
responses surprise me.  It's not like we're dealing with shared reads and
all.  Most bugs would be "I forgot to put 'synchronized' here," which ought
to be easy to spot in code inspections or even DOS FIND commands on your
code tree (looking for both "native" and "synchronized").  Not to mention
that it is hardly a requirement for Java code to be multi-threaded to start
with, though it may be prudent to make some allowance for it later on, I
suppose.

I haven't measured the performance here, but I would expect that a solution
like stored procedures (at least when there's no real data base operation
involved) could be done faster with data queues or sockets.  I don't
remember you mentioning that (perhaps you did).  That solution, too, would
alleviate concerns about synchronization.


Larry W. Loen  -   Senior Linux, Java, and iSeries Performance Analyst
                          Dept HP4, Rochester MN