TMG-L Archives

From: Darrell Martin <>
Subject: Journalling (was: Wish list item)
Date: Wed, 09 Jun 2004 05:06:03 -0500
References: <200406081058.i58AwIRA003016@lists5.rootsweb.com><40C5DC67.8000508@hockeyhockeyhockey.com><40C68C2C.20A0BEA0@sun.com>
In-Reply-To: <40C68C2C.20A0BEA0@sun.com>

At 11:03 PM 6/8/04, Jenny Joyce wrote:
>Hi everyone,
>
>[snip]
>it still doesn't seem right that TMG is so fragile. Windows is going to
>crash from time to time. That's the way of the Microsoft world. But
>this shouldn't corrupt the TMG database.

Hi, Jenny:

An operating system crash, by definition, is a
clear and present danger to any open file at
the time of the crash. TMG's database in such
a case is not "corrupted"; it is BROKEN. That
isn't TMG's fault.

>Standard database technologies allow for journalling

Not on a PC, at least not with "standard" DBMS.
Very few mainstream Windows applications have
anything like it, even if the underlying DBMS
allows for it.

> (no, that's not the
>same as incremental backups),

Correct. Journalling is a "record as you
go" technology. It is a second cousin once
removed of backups.

> so why can't TMG take advantage of this
>sort of technology.

Journalling and rollback just requires too much
in the way of resources. At work, we recently
upgraded our AS/400 (this week, IBM calls it
the "iSeries") midrange to literally triple our
throughput. The data on that machine is the
heart of our custom electronics manufacturing
business. We discussed implementing journalling,
which is built into the software we use, for
about fifteen minutes, then abandoned the idea.
We would have given back all the performance
gain we had just paid for.

> In it you still need to do regular backups. When a
>backup is done the journal is cleared. Then all transactions after the
>backup are logged to the journal. Should a restore be neccessary due to
>database corruption, you restore the backup and then reapply the
>transactions. Should you lose your hardware totally you've only got the
>last backup you did to other media, but this technology works for
>database corruptions. With this technology it IS important to backup
>(and clear journals) regularly, because otherwise there will be major
>detrimental impacts on performance, disk space requirements and time to
>restore, but still....

Your description is a good one, for the most
part. The problem is that the "but still" is
just not enough to justify the requirements to
run journalling well. Any PC built for Win 98
that was journalling TMG would probably be in
competition with a hammer and chisel on rock
for speed.

The effort required to manage a journalled
database is greater than the effort used
to maintain a good, solid, *constant* backup
procedure using TMG. Anyone who backs up at
least after every data entry session, who
takes off-PC backups frequently, and who has
a system that can identify the documents that
were used in the most recent data entry, will
be in excellent shape in the case of hardware
or operating system malfeasance.

>Jenny

In my opinion. Intelligent people of good will
may not agree.

Darrell M.


Darrell A. Martin
a native Vermonter currently in exile in Illinois
http://www.darrell-martin.net/genealogy

This thread:

• Journalling (was: Wish list item) by Darrell Martin <>