opengeometry at yahoo.ca
Thu Jan 12 04:03:17 EST 2012
On Wed, Jan 11, 2012 at 11:38:55PM -0500, Steve Izma wrote:
> So if people are dealing with structured or structurable data, why
> shouldn't they be taught (e.g.) SQL? Even better (and tied into this
> and related threads) simple scripting languages like the Shell, awk,
My opinion of SQL was never good. But it's getting better, simply
because people are more receptive to SQL. Also, from company's point of
view, they probably invested in SQL in some manner already, and don't
want to be supporting another "database".
> > Hence, my interest in your use of Python in what seems to be
> > spreadsheet related business case.
> Yes, I remember looking at this in the 80s, I think, but it was
> much too expensive. It's probably much cheaper now. But the
> concept looks good.
For Appgen... yes, it's cheap. In fact, not only is it the cheapest
ERP/Accounting solution in Linux market, but it's probably the only
solution below SAP and Oracle. Unfortunately, they missed SQL boat. If
Appgen developers early on changed their business model to run on top of
SQL database, then things would be different.
I'm trying to avoid mistakes of others. (1) If you look at other
projects that went nowhere, they failed mostly because they wrote their
own "language" or "interface", and ended up spinning wheel in order to
support it. (2) Another reason is they came up with "integrated"
solution, and ended up spinning wheel again because every company has
different business logic.
C or Python looks good. C is faster and better debugging (ie. gdb).
Python is already tapping into areas that I'm interested in tapping.
I'm also looking into Go (golang.org) which is C-like language sponsered
One interesting potential of Appgen (if re-written properly) is that it
is capable of massive parallel processing. Currently, it goes through
list of records sequentially. But, each record is complete and
independent, and they can be processed in any order. That means, they
can be processed in parallel. For example, instead of posting 10
records sequentially, you can run 10 postings simultaneously each
processing 1 record. Of course, you need to serialize at some point
when it hits the database, but that's orders of magnitude improvement.
We'll see. :-)
More information about the kwlug-disc