XML size (was: no subject)
08 Apr 2002 19:30:51 +0000
"Robert A. Uhl" <email@example.com> writes:
> > BerkeleyDB is a piece of software I wouldn't want to bet my money on,
> That's a pretty serious thing to say--I've heard nothing but good
> things about it in the past. Under what conditions were these errors
> experienced? Were SleepyCat informed? Had they any ideas? What
> version of BerkeleyDB was used?
It was BerkeleyDB 3, minor version unknown. Daily usage rate of the
database was about 15.000 read and write transactions each. A problem
that was creeping up was that for some keys, a value would be found,
but the key/value pair wouldn't turn up during iteration over the
The developer chose to use cdb instead of BerkeleyDB eventually,
instead of getting SleepyCat to fix BerkeleyDB.
I also once ported BerkeleyDB to Linux on PowerPC, and noticed that
BerkeleyDB doesn't use pthread mutexes on Linux, because Linux doesn't
support the PTHREAD_PROCESS_SHARED attribute. Instead, it defines it's
own primitives for spinlocking in assembler. Now these spinlocks
aren't shared across processes either, so what was the purpose of the
exercise? The comments don't tell.
I fully believe that BerkeleyDB works for a lot of people. I just
mis-trust it a little because of above experiences, and wouldn't let
it near really important data. Using it in Gnucash for the help
browser is perfectly fine for me, for instance, while using it for
storing my bookkeeping wouldn't. YMMV, of course.
"In my eyes it is never a crime to steal knowledge. It is a good
theft. The pirate of knowledge is a good pirate."