[Info-vax] Databases versus RMS

[Stafford Winters]

On 4/19/2012 9:29 AM, Jan-Erik Soderholm wrote:
> Paul Sture wrote 2012-04-19 16:42:
>> On Wed, 18 Apr 2012 08:32:55 -0500, Bob Koehler wrote:
>>
>>> In article<4f8dbbf4$0$24585$c3e8da3$e408f015 at news.astraweb.com>, JF
>>> Mezei<jfmezei.spamnot at vaxination.ca>  writes:
>>>>
>>>> Are database engines so fragile that a power failure will truly wreak
>>>> havok on a database requiring time consuming work and debugging ?
>>>
>>>      From time to time, things that should work, like databases, or
>>>     filesystems, don't.  Generally vendors will fix things real 
>>> quick. I
>>>     once spec'd IBM AIX in part because I assumed that IBM would 
>>> provide
>>>     a bug free filesystem.  Boy was I suprized when I had to 
>>> manually run
>>>     fsck after every power outage.  I was not suprized when the next
>>>     minor version release fixed it.
>>
>> fsck on Linux will also run every n restarts.  On my first Linux 
>> system a
>> dozen years ago, n was something like 6.  Since it was a dual boot 
>> system
>> I could easily get there several times a week.  It took so long with the
>> hardware I was using that it was time to put the kettle on. And possibly
>> make a sandwich too.
>>
>>>     "database" could also be a collection of data, without an actual
>>>     DBMS.
>>>
>>>     But it's easy to buy a DBMS that does work, and not uses it
>>>     correctly.
>>
>> I had a clear demonstration of that with some trial personal accounting
>> software for Macs, which was so slow as to be unusable.  I know I wasn't
>> running it on the latest and greatest hardware, but that same machine 
>> was
>> quite capable of running industrial strength commercial accounting
>> packages.
>>
>> It turned out that this software was using PostgreSQL, and provided 
>> dumps
>> in SQL format (plus marks for that).  OTOH inspection of the SQL 
>> revealed
>> countless indexes which only ever contained one of 2 or 3 values.
>>
>> Sorry, but I learn that indexes with only 2 values such as "M" and "F"
>> were bad news back in RMS days.  I'm pretty sure the RMS documentation
>> included that as an example of things not to do.
>
> I sure RMS doesnt handle that in a good way.

RMS does have an indexed file option referred to as a "null key" where 
the records which contain the "null value" do not appear in the index.  
ISTR the "null value" being limited to a single byte, but that could be 
used creatively, especially with more than one of these keys.