Re: Unknown SQL
Date: Sat, 21 Jul 2001 18:01:54 GMT
Message-ID: <9cv6lk$d7b$04$1_at_news.t-online.com>
Todd Gillespie wrote:
[allow me to start from back to front]
> You have proved nothing for or against any DB type -- all your technical
> arguments in this post deal with locking systems; a question wholly
> unrelated to how data is visualized.
> First you complain about learning new tools; then about the technical
> incompetence of many programmers; following up with some complaints on
> locking systems (some vague, some obvious); deciding that ignoring the
> locking system works best for you; and then wrap it all up by blaming the
> mess of SQL. What kind of magic wand would you have us buy?
> And can I have 3 wishes?
:-)
I am sorry for the impression that I wanted a discussion about database
types. If you would please be so kind to have a look at the start of this
thread:
I was drawn into this discussion with wrong statements and personal insults.
Since object databases are used by less than 1 % of information technology
(something like 250 million $ of revenues compared with multiple billions),
I am now faced with a discussion with SQL gurus with little support from my
side. So far I am still having fun at it. Please don't try too hard to
misunderstand me.
To put one point clear:
If I would be a consultant and you would ask me for a decision for a
300-concurrent-user system to be developed right now, there is no choice but
relational.
My argumentation in this thread is from the standpoint "What would be possible with an ideal object database system?" Ideally relational systems can be completely replaced. Discussing this point in detail would not make any sense since it would inevidently end up in a 100% complete specification of how such a database would have to look like.
> : What are you more effective with?
> : - a language that you have been working with for 5 years, knowing all
the
> : bugs and special tricks?
> : - a language that you use for the first time?
> : There always is a first time.
>
> Under your logic, we should cease all learning, since 'first time' costs
> are apparently not amoritzed across many projects.
Looking at the real-world-technologies behind our abstract arguments - Java and SQL - of course it makes sense today to learn both.
If an API for all SQL tasks would be available and widely adopted (total theory, but the origin of this posting) - why learn SQL?
> Fundamental fact- a programmer who learns many languages and tools will
> have a better mental toolkit than one who does not. I'm not telling you
> to code your system in a dozen languages in once; just to have people that
> know how. Why is that hard to digest?
This is not hard to digest, you simply are right.
Some real-life "but"s:
- I know some programmers that will never be hackers. They are
out-of-control with their current projects in one language. Do you suggest
that they should learn a second?
- Who pays for the education?
- How long do developers typically stay in the same company?
> Moreover, wher ein the SQL standard is it said that multi-user-problems
> are restricted to SQL alone?? User concurrency is a hard problem, it is
> found everywhere.
Very true and indeed most object databases are very immature in this area.
> : - Using Oracle in "serializable" isolation level, you are bound to get
> : bombed out with exceptions.
>
> You boggle my mind and disturb my lunch. I have no idea what you are
> talking about; I am quite happily using Oracle in SERIALIZABLE mode; what
> does 'bombed out' even mean?
Sorry, I am out of the business of using Oracle since 12 months. I have had a discussion with a colleague, whom I highly respect, last week. "NotSerializableException" might ring a bell for you. I can not reproduce the exact facts in detail, since I did not experience them myself. Posting this here was a bad idea.
Kind regards,
Carl
--- Carl Rosenberger db4o - database for objects - http://www.db4o.comReceived on Sat Jul 21 2001 - 20:01:54 CEST