From: Aaron Chang (anc327@yahoo.com)
Date: Mon May 03 2004 - 06:20:15 PDT
review 5
cse544
some of the theoretical CQ material went a little over
my head. but the
issue of containment and proving it appears to be very
important in
determining if certain types of queries can be
performed. this paper could've
really used simple diagrams to show the graphical
proofs and make points real
quickly.
the section on making queries from views (internal
resources - metadata?) was
particularly interesting. the practical consequence of
the containment test
really shows its power here. knowing what type of
subgoals are equivalent
could really be useful in streamlining queries. it's
also neat to know that
there are practical workarounds some of the
NP-complete problems CQ theorems
suggest.
this paper gets most interesting in its treatment on
IIS's - the most
practical aspect of all the prior theoretical
groundwork. the common problems of
data warehousing, especially cost and maintenance, are
the foci of developing
true metadata data integrator systems. IM and Tsimmis
are the examples in this
paper.
IM: "solution of a query is the union of all minimal
CQ's over views"
Tsimmis: object-oriented model that uses a query
language between objects (MSL)
key difference in approach: IM uses a global predicate
system, Tsmissis are more
localized one with rule expansion - my ultimate guess
is that the latter would work
more flexibly, but this might depend on the number of
"mediator" objects. which if
it still is, then we're back to square one as why it
is so hard to integrate
multiple databases. the purely global predicate system
seems more likely to work,
though it appears more rigid in adding new
datasources.
the ideas in this paper are intriguing, even if some
of the theoretical logic
underneath is a little obscure. the implications
however are very important in
my opinion.
__________________________________
Do you Yahoo!?
Win a $20,000 Career Makeover at Yahoo! HotJobs
http://hotjobs.sweepstakes.yahoo.com/careermakeover
This archive was generated by hypermail 2.1.6 : Mon May 03 2004 - 06:20:23 PDT