Re: Newbie's Oracle 9i impression: it sucks

"SQLJoe" <sqljoe_at_aol.com> wrote in message news:20020520160440.01552.00001103_at_mb-mu.aol.com...
> Mr. David Fitzjarrell, I actually don't disagree with your points. But I
think
> you can leave your hostility at your door. Yeah I bashed Oracle (in my
mind
> justifiably), but please don't overreact like the rest of the buffoons in
here.
>
> One more point, there are many of you Oracle folks who like to bash MS
SQL, and
> as a MS SQL DBA, I do NOT go ballistic when people bash MS SQL in this
room.
> You guys need to keep an open mind. Perhaps many of you should learn MS
SQL
> since it is CLEARLY taking over marketshare from Oracle in W2K
environment.
>
> Furthermore, after Windows.net 64bit OS and Intel's Itanium II comes out,
I
> believe MS SQL will chip away Oracle's marketshare as a high end RDBMS.
Don't
> go ballistic over this. Please THINK RATIONALLY before you reply.
>

Well, rationally, I think you're wrong. A revamp of an O/S and more powerful hardware does not make an rdbms 'high end'. There is a fundamental flaw with nearly all other rdbmses, as far as I'm concerned: they have a major concurrency issue.

If I update the emp table, in Oracle, you can read the emp table, and you will see the data as it was before I started updating it. That's the rollback mechanism being used to generate read-consistent images of data.

Try that in SQL Server (or Sybase etc etc), and you will discover that, unless you force the issue, your select is blocked by my update. Only when I commit can you read.

That's an architectural issue of the rdbms itself, and no amount of O/S improvement or hardware will deal with it.

Also architecturally-related is SQL Server 2000's insistence on 8Kb blocks ("pages" if you prefer). That's performance limiting. Also it has a problem with long-running transactions (as does Informix amongst others) because the same mechanism is used for both transaction recovery and transaction rollback. If your transaction takes up more than half the log, it automatically fails, because it needs as much room to roll the transaction back as it took to generate it in the first place. Oracle's separation of redo from rollback means that this simply isn't an issue for it.

The beauty of Oracle lies not in its interface, but in its fundamental architectural design.

Regards
HJR
> Jinsoo
> MCDBA, MCSD, MCSE+I
>
> >Remember, too, that your topic was, to say the least, provocative
> >(some might say assinine). I'm certain that few, if any, would call
> >it insightful since it consisted of nothing but software bashing at
> >the most base level, a technique guaranteed to generate some irate
> >responses. Then you compound such verbiage with personal attacks,
> >which is juvenile behaviour at it's worst. Anyone as 'intelligent' as
> >you claim to be would have taken the 'high road' and left the
> >grade-school antics at the door. You've, instead, chosen to fuel this
> >fire; you, therefore, reap what you sow and have no valid reason to
> >complain about the quality of the harvest.
>
>
>
>
Received on Mon May 20 2002 - 15:47:55 CDT