Re: The anatomy of plagiarism that was made by authors of "Anchor Modeling"
Date: Mon, 11 May 2015 12:14:49 -0700 (PDT)
Message-ID: <b5e9827b-4ae5-4651-b3e3-7942218f0bbc_at_googlegroups.com>
1.
"6NF" is not a solution for history because it is the result of Functional dependency
and Join Dependencies. "6NF" does not "tolerate" redundancy. But redundancy is quite
good for history. General databases in contrast to Relational Model prefer
redundancy. This is what the authors of "6NF" do not understand. This fact does not
understand also the authors of "Anchor modeling", otherwise they would not use "6NF"
for history.
2.
I have repeatedly written that 6NF does not provide any new result. Authors of 6NF does not give anything, what would a relvar R put in 6NF. There is no a procedure that relvar R, can bring to 6NF.
So 6NF is just a synonym for "atomic data structures" or for "irreducible data structures". Everyone knows that these atomic structures are very important, that is not a problem. I mean we don't need to learn 6NF for that. The problem here is how to get these atomic structures. This is what the authors of 6NF wisely kept quiet.
This decomposition into atomic structures, it is a question of how to present an
entity by using its atomic structure. So it's a fundamental question - how is built
an entity. This is the essential difference between 6NF and other NFs, and therefore
6NF must have the exact procedure that a relvar effectively leads to "6NF".
In my solution I use a completely new database theory, I called it "General
databases".
General db solves the problem, Relational Model can not resolve this problem.
This more general view of the world, Codd was not realized in RM / T and Darwen & Date in 6NF.
However the authors of the "Anchor modeling" base their work on surrogates, which are similar to Codd's surrogates and on 6NF. Moreover they put 6NF in the title of their work. Of course this is serious nonsense, because they display in the title their misunderstanding of basic things in database theory.
General db theory resolves problem. Authors of "Anchor modeling" took my results and ideas from my work and put into their data structures (see "Historized attributes and ties"). "Anchor modeling" is based on the atomic structures that are plagiarized from my work. These "their" atomic structures are the start position in the anchor modeling.
**********The first time that somebody did "decomposition" into atomic structures**** In my paper, from 2005, was shown, how the decomposition into atomic structures should be done. In my paper from 2008, was shown, how to make the decomposition on atomic structures by using concepts.
We can notice that P. Chen and Anchor modeling did not use concepts although both models are classified as the conceptual modeling. Moreover neither of the model does not define the concept. Note that even the "Relational Model" does not define the concepts. For example, E. Codd did not mention concepts, although he uses predicate calculus.
3.
The decomposition into atomic structures has to be solved in the following way:
a) The whole db theory should be divided into two parts; Simple db and General db, that is
databases that maintain the current state the databases that maintain history.
b) make the structures that allow maintenance of history
c) make the atomic structure especially for Simple db and especially for General db.
d) Add a number of specific areas. I will mention just some of them: simple key, "metadata",
"temporal data", events.
The above points a), b), c), d), very clearly show that 6NF has nothing with these above mentioned conditions.
4.
In my paper Simple Form, posted on May 15, 2006. on my website at www.dbdesign10.com
, section 4 and 5, I presented Simple Form:
4.1 Definition.
Relation schema R (K, A1, A2,...,An) is in Simple Form if R satisfies: R (K, A1, A2, ...,An) = R1 (K, A1) join R2 (K, A2), join ... join Rn (K, An) if and only if
- Key K is simple
- A1, A2,... , An are mutually independent.
In my opinion this form is better that 6NF because: a) It determine conditions under which the decomposition into atomic structures can be done. Note that 6NF does not give any conditions. b) In case when attributes are mutually independent, 6NF can not do this decomposition. c) Simple form completely describes identifiers of entities. This involve surrogates, locally defined identifiers and international-standard identifiers.
5.
Since 6NF is fundamental thing in "Anchor Modeling", then the authors of "Anchor
modeling" must present the following items and that they did not present:
(i) the definition of identity of the entities in the relational model. (ii) how they determine identity of the entities in the relational model. (iii) what is an entity in the relational model
Note that the author of "Anchor modeling" did not explain how the "entity" and "time" fit into the predicate calculus of the first order.
Note that first term in their paper is determined in "Def 1": Def 1 (Identities). Let ID be infinite set of symbols, which are used as identities.
Since they use 6NF in their model, then they should explain what are "identities" in relational model.
As I indicated at this point, the authors of "Anchor modeling" use some the most general terms from philosophy and put them in a very precise area, that is in the database theory.
Here I will give an example in which I tried to give my definition of time. My data model is defined as an event oriented model and states are defined via events, time is defined via events, so that's the event model, and not temporal etc. My model uses only two types of events. The following text is from my paper "Semantic databases and semantic machines" , look at section 6, on my website http://www.dbdesign11.com :
" 6 Events
Our model is event oriented. In terms of changes of state of entities or
relationships, we define only two events: one that creates a new state of an entity
or relationship, and one that "closes" the current state of an entity (relationship).
Thus, these two events completely determine the state of every entity or
relationship.
6.1 Definition of Time
We can also apply this approach with two events to define time. For instance, we can
understand a "second" as a set of two events which occur on the entity "clock": the
event that creates a new second and the event that closes that second.
6.2 Duration of the State of an Entity
If we wish to measure the duration of a state of an entity in seconds, then we say
its duration is equal to the corresponding number of events on the entity clock. This
way, we use events, not units of time (seconds, minutes, etc) to measure duration.
Therefore, the duration of an entity's state is relative because it is always
measured by events defining the states of another entity. Depending on the events we
use to define states in the second entity, the duration of the state of the first
entity can be different. "
6.
The main data-structure in anchor modeling is "Historized attributes," Hatt (C, D,
T), where C is surrogate key, D is an attribute and T is time. (Also see "Anchor
modeling an agile modeling technique using the normal sixth form for structurally and
temporally evolving data", Def 5.
Let entity is a car and this entity has an appropriate surrogate. The attribute of
the car is red, and T is the time from when the red color is on car.
For example (785545528897, red, 2011-11-28)
We can see that this surrogate-key is very bad, because no one knows what is this
surrogate, except the man who created it. Note that if we put the VIN here instead of
the surrogate, then everything is OK.
Here K(C,T) is a key, that is surrogate key + time. If we assume that we have a time
interval, then this structure is ruined, because we get duplicate key. For example,
let time interval is determined with the following values: DateFrom=2011-11-28 and
DateTo=2014-11-30, then
for attribute red we have:
(785545528897, red, 2011-11-28)
(785545528897, 2014-11-30, 2011-11-28)
So their main structure Hatt (C, D, T) can not support time-intervals, because they would get a duplicate key. To avoid the duplicate key authors of "Anchor modeling" invented Hatt (C, D, T) structure. This structure has only one T.
Therefore, their date is made so that it has a double meaning: (i) it means the beginning of the value of an attribute (ii) at the same time the date means termination of the previous value of this attribute.
6NF can not support time intervals in the main data structure Hatt. To support 6NF authors of "Anchor modeling" wrote the main data-structure Hatt (C, D, T) without "metadata", that is the main structure Hatt has only one "T". There is no time interval. Of course this has nothing to do with real business applications.
Note that the authors of "Anchor modeling"wrote in the paper: "Although important, the metadata is not discussed further since its use does not differ from other modeling techniques." (Look at page 2) However, this is not true because as I showed above, this model can not solve the main data-structure Hatt with metadata.
Let's try to use the "metadata" on the way that the authors of "Anchor modeling" just have described:
In this goal, we will apply the following scheme: (surrogate-key, attribute, DateFrom, DateTo, id-of-data-entry-person(or person's password))
Then we can create the following example, where the entities are cars, one attribute
of the car is color:
(785545528897, red, 2011-11-28, 9999-99-99, 122 ) this means "red" is current color.
At this moment it seems "red" color will last up to max date.
(785545528897, red, 2011-11-28, 2014-11-30, 244) DateTo was entered for "red"
color. Note that we didn't do any of updates. We just have added these above tuples.
(785545528897, blue, 2014-11-30, 9999-99-99, 201) New color (blue) starts from 30.11.2014. The operator 201 was entered this data. So in the above example, we created 3 tuples, that keep history of events. If we apply update or delete, in one of the above tuples, then we do not have history. Note that anchor modeling applies delete (I wrote about it), it means there is no history.Note that here, there is duplicate key, again.
Conclusion
With these examples I showed that "Anchor modeling" can not work history in intervals of time, because then they get a duplicate key and 6NF is inapplicable.
Of course, time intervals, often are required in real business applications. Time intervals "Anchor modeling" can not solve, because it is artificially created to meet 6NF.
Please note that in my paper from 2005, in my example, I showed solution which has one T that is plagiarized in Hatt(C, D, T). See example2 in my post from April 23, 2015, wich is in this thread.
7.
Once again, I will repeat, that authors of "anchor modeling" were working parallel
with two data models; with Relational model and with Anchor model at the same time.
After my public criticism of such a scandalous work, which won first prize at the
ER09, the authors were corrected their mistakes, so that they plagiarized again my
results, this time they plagiarized my theory of states and my theory of
identification. All this was happening publicly and globally, that is on the
Internet.
Mapping between the data model was resolved only in my papers. Please note that the
mapping between two data models, my theory of identification and my theory of states
are very important results in database theory.
Vladimir Odrljin Received on Mon May 11 2015 - 21:14:49 CEST