Claudio Valderrama C. skrev i meldingen
<88lcnv$r...@bornews.borland.com>...
Quote
>Bj?rge S?ther wrote in message <88iq72$pc...@bornews.borland.com>...
>>Claudio Valderrama C. skrev i meldingen
>><88i1sl$p...@bornews.borland.com>...
>>>Bj?rge,
>>>better later than never.
>>Oh, no problem. I'm still running the same query as I did 2 weeks
>>ago....;-)
>Nice. Can you make it run for a year and put it in the Guiness?
>:-)
Hi !
I'm fiddled with the "stopping problem", and don't know whether I should
reset the server and move on, or wait for another week or so.. One never
knows when it will be done...Would be nice with events being posted
before committing the query...;-)
Quote
>So, really, I'm grateful to exchange ideas with
>another self-taught man like you.
Well, sometimes I'm actually thinking of having a few classes in
computer science (educated sociologist & service-electrioncs), got some
big, black holes in my knowledge...
Quote
>- In ANY IB version, mixing any aggregate function with rdb$db_key will
kill
>the server immediately. In your case, I believe this suffices:
>SELECT F.RDB$DB_KEY,
> MAX(R.BET_DATO)
>FROM FAKTURA_IMP F
>You don't need to execute this, only prepare the statement and crash,
boom,
>bang. Note the sentence is ill-formed, because there's no sense in
getting
>one record-value (db_key) with a summary value (max), but the server
gives
>up before validating it.
>I assume you understand your original sentence is illegal.
Well, yes, it doesn't make sense, but a sensible version could be a
driving table returning all rows+db_key and a joined table returning
aggregate fields. The problem is that you cannot group by db_key...
Quote
>>And in case of a crash, IB seems to run into a deadlock situation,
>>locking both the server and the client. You actually have to push the
OK
>>button on the error dialog before the client knows anything about the
>>crash.
>Guesswork: there seems to be one mutex or semaphore to handle
connection and
>disconnection, so there's no way to get out of the problem in these
>deadlocks. They have been described as "system wide deadlocks". See IB
>newsgroup in mers for more details.
Are you sure about this ? I've come to that you may not do metadata
changes that would affect any active transactions.
Quote
>>I've described all the metadata "crashes" that don't cause the server
to
>>crash, just corrupts the database and produces error messages caused
by
>>vast securty holes in metadata handling.
>This is a consequence of having metadata affected by versions, too. And
when
>you change a stored proc, other users continue using the old version
until
>the last connection that used that version is terminated. Then as users
>begin reconnecting, they see the new version. This seems to be a
consequence
>of the cache for stored procedures.
Well, whatever the intention was, it doesn't do what was intended. The
worst thing I've seen, is creating a table where only 50% of the fields
were present in the created table. This happened when I did a DROP TABLE
immediately followed by a CREATE TABLE with the same name. My theory is
that poor synchronization is the reason.
Quote
>with IB Server running and the user connected to
>the Internet, it's possible for a malicious person to connect to port
>3050/tcp and use the default IB superuser to fiddle with the data.
This one is a bit scary...I'll keep it in mind...
Quote
>> I would never suggest Oracle as a platform
>>to anyone without experience, so Interbase is my choice on C/S
databases
>>for the time being.
>Maybe if you make an arrangement with some Oracle representative so
they
>install Oracle in each laptop you want to use with your programs.
>:-)
>I believe Personal Oracle is only for development, the same way the
Personal
>version of IBM's DB/2 (Universal Database) is targeted: they're hooks.
Personal Oracle is about as powerful as Paradox. No SP's, no Triggers.
It only makes sense in conjunction with some replication functionality
that I suppose Oracle is providing.
...a short, fragmented answer, to keep down quoting...;-)
--
Bjoerge Saether
Consultant / Developer
Asker, Norway
bsaether.removet...@online.no (remove the obvious)