I have noticed that the Data.fs file grows constantly. Does anyone know if the earlier part of the file is ever altered? If not, then I was thinking that it might be usable with write one media such as burning onto CD or DVD ROM. The database could be packed whenever the media was swapped out. --Sean
On Sat, Jan 24, 2004 at 06:19:01PM -0500, Sean wrote:
I have noticed that the Data.fs file grows constantly. Does anyone know if the earlier part of the file is ever altered?
IIRC, the ZODB does provide a mechanism for altering old parts of the file, but Zope does not use that mechanism, so Data.fs is effectively append-only. It is *possible* that some 3rd-party product might violate that rule, but seems unlikely. -- Paul Winkler http://www.slinkp.com Look! Up in the sky! It's THE BONOBO THE UNDERAPPRECIATED! (random hero from isometric.spaceninja.com)
--On Samstag, 24. Januar 2004 18:19 Uhr -0500 Sean <whysean@softhome.net> wrote:
I have noticed that the Data.fs file grows constantly. Does anyone know if the earlier part of the file is ever altered?
Data is always appended to the Data.fs file. What's the problem with setting up a cron job that packs the ZODB every night or so?
If not, then I was thinking that it might be usable with write one media such as burning onto CD or DVD ROM. The database could be packed whenever the media was swapped out.
Not a brilliant idea... -aj
Sean wrote at 2004-1-24 18:19 -0500:
I have noticed that the Data.fs file grows constantly. Does anyone know if the earlier part of the file is ever altered?
It is routinely: When "FileStorage" starts to write a transaction record to the storage file, it marks the transaction status as "checkpoint" (status == "C"). After the complete transaction record is written, the transaction status is changed to " ". Thus, you cannot place the storage file on a write once medium (unless you change the "FileStorage" code). -- Dieter
Dieter, Thanks for the info! That looks like something small enough that it could be changed. There are a lot of applications that use write once tamper resistant technologies to guarantees data security is linked to physical security. With SQL databases, this is mostly done with a log file containing all the SQL commands, and is really only useful for checking later when one suspects that security has been compromised, or for rebuilding when there is file corruption. It is intriguing that Zope is so close to being able to use the actual live database file in this way. If this could be made to work, one might be able to guarantee against both illicit data alteration and accidental data corruption of the live data, to the tolerance levels of one's physical security. --Sean Hastings
-----Original Message----- From: Dieter Maurer [mailto:dieter@handshake.de] Sent: Sunday, January 25, 2004 6:24 PM To: Sean Cc: zope@zope.org Subject: Re: [Zope] Write Once Media with Data.fs?
Sean wrote at 2004-1-24 18:19 -0500:
I have noticed that the Data.fs file grows constantly. Does anyone know if the earlier part of the file is ever altered?
It is routinely:
When "FileStorage" starts to write a transaction record to the storage file, it marks the transaction status as "checkpoint" (status == "C").
After the complete transaction record is written, the transaction status is changed to " ".
Thus, you cannot place the storage file on a write once medium (unless you change the "FileStorage" code).
-- Dieter
participants (4)
-
Andreas Jung -
Dieter Maurer -
Paul Winkler -
Sean