Why researchers should look at product features for related work
Many presentations at events such as VLDB, SIGMOD or XSym present interesting and novel ideas. As is commonly required from academic papers, the presentation of these ideas needs to include a comparison with related work. This has several benefits:
- It makes sure that the presenters check that their idea is indeed novel
- It gives readers a way to quickly understand the new idea based on the differences to old ideas
- It gives readers the ability to learn about other work in the same area
- It provides the researchers baselines to judge their own work against
Normally, papers that get published have a good amount of comparison with related academic work. However, in some areas, one needs to look beyond the related academic work and look at what has been done by the industry in released products.
A case in point is a presentation on querying XML data using SQL. The presenter talked about mapping SQL into XQuery and finally presented a prototype that extended a competitors component to map XML into tables that is very close to our OpenXML mechanism. E.g., his paper can be described in the following way in SQLServer 2000/2005 (similar functionality will be/is available in DB2 and Oracle):
select POId, PODate, …
from OpenXML(@XMLdoc, '//PurchaseOrder')
with (POId nvarchar(10) '@id', PODate datetime 'orderdate', ...)
Now, I am quite happy to see somebody else provide a similar approach to query XML in a relational context using SQL as we provided already in SQL Server 2000 (it validates our approach), but if the authors and/or the reviewers would have looked at the functionality provided by us and our competitors, the available related work would have shown that the idea of generating tables out of XML and then using SQL to query them was not really novel. And maybe the current implementations would have given the authors ideas about improving what is available: e.g., how to push SQL predicates into the XQuery/XPath predicates for better performance…
Often researchers complain about the lack of availability of information about what products really do (the marketing material does not cut it). This used to be a valid criticism when you needed to buy expensive manual libraries. But in today's world, where technical websites provide a vast array of technical resources about features and people from the product teams are publishing at conferences (for example, Shankar presented our indexing mechanism on the XML datatype at VLDB) and are otherwise accessible, I don't buy that argument. Especially not in cases such as the one above. There may be some internal details that companies will not disclose, but a general technical description of publicly accessible features is always available (and sometimes even more).
If you are a researcher and want to know more about the XML functionality that we provide, feel free to send me a note or check out the MSDN resources.