"The OverRelational Manifeso". The answer on x's question (RM/T and "TheORM")
Date: 4 May 2006 00:48:26 -0700
Message-ID: <1146728906.485024.206200_at_j33g2000cwa.googlegroups.com>
In my previous topic "x" asked me what is relationship between "The OverRelational Manifeso"'s approach and RM/T's approach. http://groups.google.com/group/comp.databases.theory/tree/browse_frm/thread/0194f68088b82cd4/675529ca9d919267?rnum=11&hl=en&_done=%2Fgroup%2Fcomp.databases.theory%2Fbrowse_frm%2Fthread%2F0194f68088b82cd4%3Ftvc%3D1%26hl%3Den%26#doc_e0d96072d02facf7
IMHO next idea of RM/T is true - Meaningful units of information larger then simple n-ary relations exist (Mr. Codd names they as "molecular semantic unit") - and "The OverRelationаl Manifesto" bases on this idea too.
RM/T classifies kinds of entities and offers a number (ten) of types of relations and, also, special constarins and operations on these types of relations. IMHO it looks like a straining of single relations on entities. This means that we have to find the entitites, which are simple enough, to be represented in single relations.
For example describing some "Shipment" RM/T will find entities "Header"
and "Items" of course, becouse these entitites is enough simple to be
described with simple relation each. In final we get four relations
(E-relation 'Shipment", P-relation "Shipment_Header", characteristic
relatipoon "Items" and CG-relation "Header_Item"). Together all these
relations present meaningfull units of information, which can be named
"Shipment", and this unit of information is larger than simple
relations. (By the way - is "Shipment" entity too? :)
I will not go into the details of this schema. Some difference (big
enough) between it and the "TheORM"'s R-projection schema exists and
can be discussed. But this is not the main difference between RM/T and
"TheORM".
- The fact is that RM/T says nothing about how we can realise this ideas. For me it looks like we have to do it all (create these relations and, also, special constarin and operation) manually but this is difficult really. "TheORM" offers the approach allows create a system, which will performes all this tasks (Once again – "TheORM" doesn't offer new model – RMD is only used data model, all other question are about principles of system). This system creates SET of relations using description of this moleculare semantic unit, which are more complex than simple (and single) relations.
- At that "TheORM" says nothing about kinds of entities, which can
exist in entreprise and don't find simple ones. On the contrary - it's
only assumption that the meaningful units of information can be really
complex – as complex as relational DATA BASE (= SET of relations)
can. The system, which "TheORM" speaks about, is able to describe such
complex structures and, then, translate this description into
relational schema. I don't believe that some universal infological
model can be created. I know person who believes that only binary
relationships exist and he defends this point of view violently, And I
know other persons who have very intresting, sometimes strange but
really different approaches on creating of enterprise description. So,
"The ORM" doesn't try to classify entities, attributes, relationships
ets, what can exist in the enterprise. Instead of this it's found possibility to describe more complex "molecular semantic units" which it calls just as "object" (becouse it looks like the ones in usual OO-languages) and to manipulate with corresponding variables. "THeORM" supposes that developers themself will decide how this possibility can be used in a system.
Of course, when systen translate such complex description, it gets and
can save information on moleculare (overrelational) semantics – I
mean it as a molecular schema which combines a number of atomic
relations schemas. Really this information links definition of
molecular semantic units (let us call them MSU futher…. or let me
call it just object:) and relations schemas. And IMHO next step is
logical enough – we can use this information to translate any
commands on MSUs variables (IMHO objects) to commands on relations.
"TheORM" writes a lot about such translation. The result looks like
MSU(O)-Oriented :) managemnt system on relational database.
My opponents cannot understand one simple idea. Speaking on possible
relationship between OO-world and RM-world "TheORM" supposes that
object's counterpart in relation world is relational data base (This is
"TheORM"'s third answer on TTM initial questions I spoke about)
http://groups.google.com/group/comp.databases.theory/tree/browse_frm/thread/0194f68088b82cd4/675529ca9d919267?rnum=11&hl=en&_done=%2Fgroup%2Fcomp.databases.theory%2Fbrowse_frm%2Fthread%2F0194f68088b82cd4%3Ftvc%3D1%26hl%3Den%26#doc_c3f45205c2a91c8a.
This is main requirement allowing organize such a translation.
Received on Thu May 04 2006 - 09:48:26 CEST