Indexing problems

Okay, to fill you in on the testin with gdbm. I tested this today.
What I did was build a version of perl using the gdbm libraires.
The basic results are that it indexed fine and the size of the dbm
files were nice and small. Slightly smaller than the actual database. 
There were no problems whatsoever. However, here is the downside....
The index itself took almost three times as long to build. I currently
do not have more time to do any real testing at the moment but this
does highlight a basic bug in the standard SUN dbm somewhere. For now
we will go with Martens workaround and if MErit folks have more time
to spend looking at the pros and cons of other dbm implementations
we'd be very interested.

		--Tony.

 Marten Terpstra <Marten.Terpstra at ripe.net> writes:
  * 
  * Sigh, I am not sure I like dbm ... ;-) I found a cure I think. Up till
  * now I went to my escape mechanism for large buckets whenever the
  * current bucketsize and whatever I wanted to add exceeded 1000 chars
  * (because I assumed the bucket max to be 1024). Then I suddenly noticed
  * that the funny dbm bevaiour happened when the old bucket and to be
  * added value was *exactly* 1000 chars. I changed my test to 950 chars
  * and voila it worked ....
  * 
  * Could the MERIT folks please change the following and see if they can
  * index the prdb?
  * 
  * in cldb.pl routine setmspnxl() change:
  * 
  *     if (length($value) + length($addvalue) > 1000) {
  * 
  * to
  * 
  *     if (length($value) + length($addvalue) > 950) {
  * 
  * 
  * install and run an index. It worked for me on prdb ....
  * 
  * -MT



-------- Logged at  Wed Nov 2 11:36:20 MET 1994  ---------