Inefficiency comments

[Mark.Koennecke at psi.ch]

        More comments on efficiency,

        Ray asked for more comments, so here they are:

        - Who HAS expertly stated or proven that vData are more efficient then
          SDS? I looked at the vdata API and found a lot of calls to make.
          For instance a 2d detector would have to be read record by record.
          Who wants this?
        - Aren't SDS the way to deal with 3d TOF datasets? How would those
          fit in a vData? 
        - To Chris: a NeXus file (IMHO) resembles more an object data base
          file then a relational data base file.
        - I agree, that HDF is inefficient in storing small data items. Even 
          the HDF-docs say so. However, how many of those do we have? Most
          of them are static: this gives us some 20-30 at maximum in the 
          first entry vGroup. Further entries are supposed to link to
          unchanged data items from the first entry. Our data will be 1d-arrays
          (scans), 2d arrays (psd dumps) or even 3d arrays for TOF augmented
          by a few number which might denote varied environment conditions.

        At the bottomline I'am reluctant to add vData (and much more complexity)
        to the NeXus API as long as nobody has proven with a profiler output
        that his program spends most of its time writing NeXus. And it is 
        proven that vData are so much more efficient. Even then (case 1, 
        profiler) I would rather prefer to investigate if our data can not be 
        structured more efficiently.    
     
                                                        Mark
 
	
--------------------------------------------------------------------------------
Return-Path: <Mark.Koennecke at psi.ch>
Received: from dns2.anl.gov by ipns.pns.anl.gov (MX V4.2 AXP) with SMTP; Tue,
          09 Sep 1997 07:22:51 CST
Received: from osborn.msd.anl.gov (osborn.msd.anl.gov [146.139.244.39]) by
          dns2.anl.gov (8.6.11/8.6.11) with ESMTP id HAA00200 for
          <ROsborn at anl.gov>; Tue, 9 Sep 1997 07:22:49 -0500
Received: from dns2.anl.gov (146.139.254.3) by osborn.msd.anl.gov with SMTP
          (Eudora Internet Mail Server 1.1.2); Tue, 9 Sep 1997 07:22:27 -0500
Received: from pss201.psi.ch (pss201.psi.ch [129.129.40.201]) by dns2.anl.gov
          (8.6.11/8.6.11) with SMTP id HAA00188 for <NAPI at anl.gov>; Tue, 9 Sep
          1997 07:22:17 -0500
Received: from psiclb.psi.ch by pss201.psi.ch; Tue, 9 Sep 97 14:25:26 +0200
Message-ID: <97090914220104 at psiclb.psi.ch>