File store (was Re: Salutations)

Christopher Browne cbbrowne@hex.net
Mon, 11 Dec 2000 08:57:30 -0600


On Mon, 11 Dec 2000 08:13:59 CST, the world broke into rejoicing as
grib@gnumatic.com (Bill Gribble)  said:
> On Mon, Dec 11, 2000 at 01:57:31AM +0000, Al Snell wrote:
> > Archives in SQL databases are useful for data mining and OLAP, but
> > are more costly to store - anm archive file can be dumped onto tape
> > or just deleted by the user at their convenience...
> 
> Keep in mind that in the financial context, you are not-infrequently
> asked to do simple "data mining" in the form of any report run over 2,
> 5, or 10 years of your history.  The problem of multiyear reports is
> the biggest roadblock to book closing ATM.

I'd suggest that "roadblock" is not quite the right word.

The "need to do multiyear reports" represents the biggest reason to Not
Bother Closing The Books.

On the other hand, "inability to scale, whether due to limited disk, RAM,
inadequacy of data structures and algorithms, or Sheer Volume of Data"
represents the forcible reason to Yes, Indeed, Close The Books Now.

Also in there is the auditing consideration of "If the period is closed,
and people can't post there, then the past periods of data become `stable.'"
Which is implementable by storing information on what periods of time are
active and which have been finalized, and then refusing postings into those
periods that have been finalized.

There's three reasons here both pro- and con-; they're of different
natures, and do not _force_ any particular implementation of solutions
to the constraints they introduce.  
--
(concatenate 'string "cbbrowne" "@ntlug.org") <http://www.hex.net/~cbbrowne/>
"take USABLE from UNSTABLE and you get NT"