Re: Yarn 0.91 expire -r bug

David Meade (dmeade@netcom.com)
Sat, 29 Jun 1996 11:02:33 -0700

In article <xlB1xsw3BIFX091yn@iglou.com>, mtj@iglou.com (Michael Jones) wrote:

First of all, thanks to Michael and Mark for their answers. I am CCing
Chin Huang my response in addition.

>This may very well be a bug. However, you were aware, were you not,
>that .9x does not reclaim space on the drive unless the expired
>articles occur at the end of the newsbase? In this case, it looks
>like maybe 11.5MB and 5.5MB, respectively, were expired at the end of
>the newsbase, resulting in truncation of the base, while the rest of
>the expired articles occured elsewhere in the newsbase, resulting in
>them being used on the next import.

Yes, I assumed that that was the cause before I posted my msg. The
problem is that the deleted bytes from the expire don't quite match up
to that explaination. The two expires were run without any intervening
imports. Also the deletions were run against newsgroups that were
in the middle of the imported packets, and this doesn't quite match up
to your explaination. I would have expected the news.dat to have
remained the same size using your explaination.

>If this is the case, then it's not a bug, it's the way Chin is
>handling the newsbase from .9x up (I presume he won't be going back to
>the old way -- just as well, because I prefer it this way).

I disagree, it is a bug (a design flaw). I understand the logic behind
it though. By not reclaiming space, appreciable time is saved in the
expire process. Which is a very good thing for most situations.

Assuming your explaination is correct, what is needed is a seperate
"database compression" program that will go thru and reclain empty space
in the database. In my opinion there are two classes of users for Yarn:
"expire -o" users or the
"expire -r" users
I am an "expire -r" user, this means I use the database as a database
and regularly data mine it, both for short term and long term projects.
This means I never expire by date, and keep articles around for a long
time, years sometimes (yes, I know about folders). This means that I
will be stuck with a huge, mostly empty database. This is not a good
thing.

As for tne new database structure, I have mixed feelings about it.
the good thing is that CVT09 picked up the orphaned news files and
seemed to handle them, the bad is that incremental backups (between
full backups) are now much bigger.

Thanks,

David

--
 David Meade                        Internet=dmeade@netcom.com
 Oakland, CA                                 dmeade@slip.net