RE: Thats all folks!

On Tue, 17 Dec 2002, Joe Pluta wrote:

> > But data loss absolutely never happens.  Unscheduled
> > downtime (excepting power outages that take out both machines) does not
> > happen.
>
> Both happen on Linux machines.  On Unix machines I have seen various
> databases (in order, Informix, Sybase and least frequently Oracle) go
> casters up and require lengthy rebuilds.  I think it depends to a large
> extent on system load, and (one would hope) it happens less frequently now.

I was speaking from my experience, which I didn't say before.  In my
experience, I have never had data loss.  I have never had unscheduled
downtime excepting power outages.  I have read on various mailing lists
that others have lost data.  Not me - I guess I'm just lucky :)

I have experienced database corruption due to sudden power loss on the
AS/400.  By this I mean corrupted logical files.  The physical files were
fine.  This has happened several times, most recently two weeks ago.  When
this happens it does require (sometimes lengthy) rebuilds.

I think it is important to consider the different uses of these machines.
I don't run large relational databases on the unix machines, we have
AS/400s for that.  The unix boxen perform other mission-critical (I love
that hype word) tasks that make large demands of the system.  The AS/400
runs database applications that make large demands of the system.  So I
can't really test logical file corruption on unix since there aren't any.
But both OSes are extremely stable.  Neither is more or less stable than
the other.

> I knew this was going to go this route.  I talked about rate of change -
> while other platforms had these things, the AS/400 didn't.  I thought that
> was pretty clear.  But I knew someone was going to jump in and miss the
> point and state the obvious - that other platforms already have these
> things.
>
> I honestly don't know how to write any clearer, but I will try to restate my
> point, focusing on the real issue: the rate of change on the AS/400 has been
> far higher than any other platform the last (N) years.  What this has done
> is open the platform to all the things that everybody complains the AS/400
> is lacking, while keeping legacy investment intact.  No other platform has
> managed that.

No, I didn't miss your point.  But all those things you mentioned proved
that the rate of change on other platforms must be the same as on the
AS/400.  One platform gets Apache, the AS/400 gets Apache - they both
change the same.  On some platforms Apache has changed considerably from
what is on the AS/400 so the rate of change on other platforms continues
at about the same pace as the AS/400.

But most interestingly is the idea that the AS/400 changes faster than ony
other platform over the last 10 years.  That's just plain wrong.  11 years
ago linux didn't even exist.  The GNU project didn't have a whole lot.
Most PCs run running DOS 5.0.  A System 36 looked much like an AS/400
today.  The idea that an OS which had much of its groundwork laid in 1991
has changed faster than on OS which was hatched on an undergrad's PC in
1991 is ludicrous.

By the way, doesn't that paragraph make people happy to see how much good
work was done by IBM way back in 1991?  The 36 was a pain but the
underlying architecture was good.

> > If these unique components fit your needs (which they do for many)
> > then it is a better machine than the others.
>
> It's uniquenesses are what make it best suited for business applications.

And this was better stated by Scott than me.  The strengths of the AS/400
aren't really stability, performance, and scalability vs. other platforms
but rather the unique things (like I said: RPG, DDS, etc.).  Other
platforms are just as stable, perform as well or better, and are more
scalable.  So in those areas the AS/400 doesn't hold an advantage.  But it
certainly does in others.  It is focusing on the unique points that
determine whether it better suits one's needs.

James Rich