SET RMS question

[Willem Grooters]

A quick question on SET SRM_DEFAULT
Specified /BUFFERS=255, but without /INDEXED, would that cause corruption when copying/analyzing/converting indexed-sequential files?
FYI:
The primary key of a file consists of a letter and 6 digits - the letter may be uppercase or lowercase - the difference is crucial. But some way, after SET RMS, the file seems to have lost it's lower-case lettered keys.

[Hein van den Heuvel]

Setting a large number of local buffers of course should not cause corruption. If you believe it does, then be sure to report it to HP support with the exact (up to date?) system version and hopefully a reproducer of some sort. I'm not aware of problems in this space.

Please note however that setting a large number of local buffers is entirely useless and potentially counter productive for most applications.

If at all possible be sure to evaluate a progression like 4, 16, 64,... buffers.
I do not expect any improvement beyone 8 buffers and an increased CPU time and lock usage beyond 30 buffers (approx). This woudl depend on you exact system (CPU speed, memory speed) and would nto be VMS/RMS version dependend.
If you decided to experiment with this, then kindly report back the results?!

The problem with local buffers is that RMS maintains them in a simple sorted (hmmm, I shoudl check that) linked list and has to walk that list every time again, where really (notably for sequential files) you do not expect to have more that 1 active buffers in general and a few more during the read/write times.

Convert (for output) and ANAL/RMS use BLOCK IO, so they do not use the standard buffers at all.

Check with ANAL/SYS... SET PROC... SHOW PROC/RMS=(FAB,BDBSUM)

It is really hard to imagine how it would impact lower/uppercase letters. The only time those sould have an impact is in the indexed file sorting order. The key values with lowercase chars will normally come first in the indexed file, unless the key has a specially COLLATING SEQUENCE key type.

For better help, please attach a (.txt) file with pertinent details (dump/reco, anal/rms/fdl...)

Hope this helps
Groetjes (uit Amsterdam deze week)
Hein.

[Doug Phillips]

> lowercase chars will normally come first ...

Pardon my insubordination (;-) -- I've never used lowercase letters in the key before -- but since uppercase letters have a lower ASCII value than lowercase, wouldn't uppercase come first?

[Hein van den Heuvel]

Oops, dunno why I wrote that the wrong way 'round. Of course uppercase sorts before lowercase in the normal ascii table.

Hein.

[Willem Grooters]

Curiosity has been satisfied.
As expected, the problem that triggered the question was not at all RMS related.