Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.misc -> Re: Access to heterogeneous data sources - XML

Re: Access to heterogeneous data sources - XML

From: Andrea M. Segovia <andrea_at_mun.ca>
Date: 15 Mar 2004 11:50:50 -0800
Message-ID: <7589088f.0403151150.279f7bd9@posting.google.com>


.....excerpts deleted...
> Well your "technically savvy" clients aren't that savvy or that technical.
> Sounds like they are good at throwing around tlas. (good if you are in
> marketing, not if you actually have to deliver something)
>
> There is not simple solution. Just because you can retrieve the meta data
> doesn't mean that it is useful. Just because a table is called fish and a
> field is called color in two different systems doesn't mean you can merge
> one into another. The color field might be a character in one system with
> the color (red, yellow,...) the other system might be an id that relates to
> a color look up table. (or it might be a numeric triple RGB) So you
> actually have to know what each table, field, and relationships are in each
> system before you could even hope to "query them in real time". Sure I can
> read the data dictionary from Access, or MS SQLServer, or Oracle via ODBC,
> but it just doesn't tell me much other than the names of the tables, fields,
> and data types. It does not hold sufficient information to tell you what
> the tables are for or what the columns are for. What is the business need
> to query them in real time? Do the number of fish suddenly change? I think
> your "technically savvy" clients are rather technically naive.
>
> Your big problem is how to extract and transform the information. One
> poster suggested web services. That's fine, it you can install a program on
> each remote site that knows how to read that specific database and present
> the data in some agreed upon method. (the program on each machine would be
> doing the ETL on demand, but each program would have to know the particular
> meaning of the local schema and translate that into the schema that you
> want.) So you still have your basic problem. The technology doesn't
> magically take that away.
>
> Google the Internet for Fred Brooks No Silver Bullet or read his book The
> mythical Man Month. Very relevant in the development of software.
> Jim

You are basically confirming what I thought....

However, in all fairness to the clients, the application that they are looking at is fairly specialized (stock assessment for the most part, I believe) and the information that would be used as input for these models would generally be somewhat standard, thus the data would likely be similar anyway. So some of the issues regarding the structure and semantics of the data may not arise.

Unfortunately, there is no way to persuade clients who want to use a grenade instead of a fly-swatter to kill a fly because the grenade is "cooler" and "newer technology" - common sense arguments do not always hold sway.

It seems to me that the data should be gathered in a data warehouse and then applied to the models....the first phase should deal with issues related to differences of data....and real-time access should not be required...

Sadly for me, I am just the tech person that gets caught in the middle ...between the project manager promising the world and the client asking for it...

Andrea Received on Mon Mar 15 2004 - 13:50:50 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US