However, when the metadata management is necessary, there are still lots of questions:
how to design, store, and display metadata. Some high-end ETL tools have a
support for storing metadata more or less comprehensively, although rarely interchangeable
with others (limits in metadata federations).
We have not got a great expectation in these techniques, we believe that a good,
well-structured traditional documentation may constitute a valid metadata support;
the human language has the flexibility, the undertones, and the richness to depict and
contextualize any kind of information. The main problem is that paper documentation
is not linked with the data scheme, and also stored metadata need a human
intervention to reflect changes. This is a problem concerning the organization of the
???productive process??? more than an IT problem. Accordingly, it could be extremely
useful to have something like a structured language for metadata representation.
Another aspect of interest in ETL is cleaning, that is, the ability to check, filter, and
correct mistakes or corruption that can be found in the source data. A typical case
in which a cleaning process is mandatory is in the address (location) processing:
???street,??? ???st,??? ???st.,??? and so forth; all indicate the word ???street??? and an automatic
process must be able to recognize it.
Pages:
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194