[V4RB] Performances variations

[Stan Busk]

Hi,

I have a strange problem. My application loads from the internet dozens 
of thousands of messages I store to a Valentina Database. Browsing thru 
records is fast. I quit the application and launch it back and it 
remains fast but as soon as I  shut down the computer and start it back 
browsing is much much slower, including to show single records I get 
thru SQL queries. I am really puzzled. I have repeated the test several 
times and it always behave the same way, as soon as your restart the 
computer browsing is much slower, the application becomes sluggish... 
only workaround is deleting all the database files and downloading all 
the thousands records again, it becomes fast as hell...including 
quitting and starting the application back... I have tried with several 
RB versions (oficial and last beta), on a G5 Dual, a G4 and an iBook. 
Is there something I should consider? By the way if for example I have 
20,000 records and it is sluggish because I just restarted and I 
download 10,000 more records it becomes fast!... including on first 
20,000 records!...It  seems nonsense... I use a cache of 50*1024*1024 
(changing it seems useless) and a segment size of 32*1024 (that I 
believe it is the default) and last V4RB 1.10. Question is what can I 
do to maintain always the same speed? The table has the following 
structure:

   mFrom         = new VText     ("Article_From", 128, "ASCII", -1)
   mDateSent     = new VDateTime ("Article_DateSent", kV_Indexed)
   mDateReceived = new VDateTime ("Article_DateReceived")
   mSubject      = new VText     ("Article_Subject", 128, "ASCII", -1)
   mLines        = new VLong     ("Article_Lines")
   mNumber       = new VLong     ("Article_Num" )
   mStatus       = new VvarChar  ("Article_Status", 32, "ASCII", 
kV_Indexed)
   mReferences   = new VText     ("Article_References", 128, "ASCII", 
kV_IndexByWords)
   mMessageID    = new VvarChar  ("Article_MessageID", 128, "ASCII", 
kV_Indexed + kV_Unique)
   mNewsgroups   = new VText     ("Article_Newsgroups",  128, "ASCII", 
kV_IndexByWords)
   mBody         = new VText     ("Article_Body", 128, "ASCII", -1)
   mHeader       = new VText     ("Article_Header", 128, "ASCII", -1)

The reason why I use VText is because 
mFrom,mSubject,mReferences,mNewsgroups fields  can be from 1 char up to 
998 and mBody and mHeader has no limit.

TIA,
~/Stan