 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| HomeDiscussion Groups Database Servers DB2InformixIngresMS SQLOraclePervasive.SQLPostgreSQLProgressSybaseDesktop Databases FileMakerFoxProMS AccessParadoxGeneral General DB TopicsDatabase TheoryRelated Topics Java Development.NET DevelopmentVB DevelopmentMore Topics ... |
Database Forum / General DB Topics / DB Theory / June 2005Poll: What percentage advantage are RDBMS vendors taking of the RM?
> Afredo elsewhere responded to: >> My point, which you appear to agree with, is that the vendors >> are indeed taking some advantage from the RM. I am not at >> this point concerned with 'How Much', only that it is not null. > Of course it is not null, that is evident and nobody said the > contrary. You don't need to understand the Relational Model in order > to take some advantage on it. How much now, as a percentage out of 100, are contributors to this newsgroup prepared to estimate the "realisation" of the relational model of data within the SQL-DBMS vendors (ie: IBM - DB2, Oracle, MS - SQL Server) The rating of 100 equates to full realisation of the RM. The rating of 50 equates to half realisation of the RM. The rating of 10 equates to 10% realisation of the RM. The rating of 0 equates to zero realisation of the RM My estimate/opinion is around 80% give or take 10%. What is your estimate and/or opinion?  SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au/software >>Afredo elsewhere responded to: > [quoted text clipped - 19 lines] > > What is your estimate and/or opinion? there seem to be two questions here. my answer to the first one is that i am zero percent prepared to answer the second one, let alone willing. if you really, truly want some more numbers, save us all some time and make some up. that would be as useful as asking the second question. p >> The rating of 100 equates to full realisation of the RM. >> The rating of 50 equates to half realisation of the RM. [quoted text clipped - 7 lines] > there seem to be two questions here. my answer to the first one is that i > am zero percent prepared to answer the second one, let alone willing. There is one question. In what percentage is the RM "realised" in the major SQL-RDBMS products? I am interested to try and evaluate how people think about the "realisation" of the RM in today's industry. There are a widely ranging series of perspectives, and I am attempting to gauge this series. Do you have a problem with this question? SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au What you have asked is akin to, "How big is your database?" It is not a valid question in that it does not indicate what parameters you expect to be used. For example, do you mean a percent of "expected" features? Are they weighted by importance? Important to whom? For what purpose (E.g., OLTP, OLAP?) Why would anyone here be able to give a more valid answer than the thousands of practitioners alsewhere on the net? Isn't it best answered through careful unbiased research of the subject? As asked, it borders on being a silly question. > >> The rating of 100 equates to full realisation of the RM. > >> The rating of 50 equates to half realisation of the RM. [quoted text clipped - 18 lines] > > Do you have a problem with this question? > What you have asked is akin to, "How big is your database?" It is not a > valid question in that it does not indicate what parameters you expect to > be > used. For example, do you mean a percent of "expected" features? Are they > weighted by importance? Important to whom? For what purpose (E.g., OLTP, > OLAP?) What I am trying to ask is "How *relational* are DB2, Oracle and SQL Server according to the principles of the relational model? People assert that these SQL-DBMS are "the best we have at the moment" and that they are not ("Fully") relational. I want to understand just HOW MUCH they are not fully relational, as a ball-park rough-as-guts estimate. Is it 10% or is it 90%. ? > Why would anyone here be able to give a more valid answer than the > thousands of practitioners alsewhere on the net? Isn't it best answered > through careful unbiased research of the subject? IMO this question is best answered by those who have a good understanding of the relational model ONLY ! Hence the Q here. >As asked, it borders on > being a silly question. Well, hopefully I have restated the question in less border-line terms. Thanks for the prompt. Does this make more sense? >> >> The rating of 100 equates to full realisation of the RM. >> >> The rating of 50 equates to half realisation of the RM. [quoted text clipped - 20 lines] >> >> Do you have a problem with this question? > What I am trying to ask is "How *relational* are DB2, Oracle and SQL Server > according to the principles of the relational model? [quoted text clipped - 4 lines] > > Is it 10% or is it 90%. ? Well here's a list of Codd's 12 rules, which are general guidelines for a DBMS to be thought of as relational: http://www.wildewood.co.uk/comp/more/codds_rules.html I'd say most SQL-DBMSs are inspired by this, although some of the rules clearly aren't followed. But we're getting there slowly. It's easy to criticise DBMS designers for not building a fully-relational product, but I think the practical problems are bigger than we might think. It's a serious level of abstraction, especially when you consider that many DBMSs were started years ago when computer were many orders of magnitude less powerful than they are today. I'll give them 8 out of 10 for effort. :) Paul. >Well here's a list of Codd's 12 rules, which are general guidelines for >a DBMS to be thought of as relational: Codd's 12 rules were outdated by Codd's 333 rules, and Codd's 333 rules were outdated by The Third Manifesto. >It's easy to criticise DBMS designers for not building a >fully-relational product, but I think the practical problems are bigger >than we might think. It's a serious level of abstraction, especially >when you consider that many DBMSs were started years ago when computer >were many orders of magnitude less powerful than they are today. Excuses :) >I'll give them 8 out of 10 for effort. :) I'll give 4 for results. Regards >> What I am trying to ask is "How *relational* are DB2, Oracle and SQL >> Server [quoted text clipped - 23 lines] > > I'll give them 8 out of 10 for effort. :) Thanks for the "quantification". So we have two 8's, a 4 (Alfredo) and alot of nulls. SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au [snip] >Well here's a list of Codd's 12 rules, which are general guidelines for >a DBMS to be thought of as relational: [quoted text clipped - 3 lines] >I'd say most SQL-DBMSs are inspired by this, although some of the rules >clearly aren't followed. But we're getting there slowly. I really wonder about that. The XML bandwagon is one of my concerns. >It's easy to criticise DBMS designers for not building a >fully-relational product, but I think the practical problems are bigger >than we might think. It's a serious level of abstraction, especially >when you consider that many DBMSs were started years ago when computer >were many orders of magnitude less powerful than they are today. You mean when computers were used in the effort to send men to the moon? >I'll give them 8 out of 10 for effort. :) Sincerely, Gene Wirchenko > [snip] > [quoted text clipped - 8 lines] > I really wonder about that. The XML bandwagon is one of my > concerns. It's especially bad because XML attracts the many developers who seem to have no knowledge of SQL (much less relational). Why, I don't know... having worked with DBMSs and XML, doing both manual "mapping" code (JDBC/SAX/DOM) and "mapping frameworks" (Hibernate/Digester/JAXB), there's really no benefit whatsoever to XML in that regard. And its problems are described well elsewhere. Frankly, no JDBC code, however poor, can ever match the horrors of traversing an XML tree. Even bad JDBC can be refactored fairly easily; untangling XML manipulation code should have been another labor of Hercules. I'd rather clean out the stable. > >It's easy to criticise DBMS designers for not building a > >fully-relational product, but I think the practical problems are bigger [quoted text clipped - 4 lines] > You mean when computers were used in the effort to send men to > the moon? Excellent! - erk > > What you have asked is akin to, "How big is your database?" It is not a > > valid question in that it does not indicate what parameters you expect to [quoted text clipped - 24 lines] > Well, hopefully I have restated the question in less border-line > terms. Thanks for the prompt. Does this make more sense? Yes, it is clearer now. Another thought, though: Is a 100% relational RDBMS a good thing, or are some modifications a good idea after all? > >> >> The rating of 100 equates to full realisation of the RM. > >> >> The rating of 50 equates to half realisation of the RM. [quoted text clipped - 20 lines] > >> > >> Do you have a problem with this question? > Yes, it is clearer now. Another thought, though: Is a 100% relational > RDBMS > a good thing, or are some modifications a good idea after all? When you start getting close to 100% I believe a range of issues will emerge that effectively resolve to different interpretations of a number of principles within the RM. For example, Codd and Date's different approach to the way null values should be handled by the RDBMS. Is there one model, or two, and/or which path gets to *the* 100% ? Setting aside these more contraversial issues, the over-riding principle (as it appears to me) is that you need a very good reason to diverge from the relational practice. SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au >> Afredo elsewhere responded to: > [quoted text clipped - 19 lines] > > What is your estimate and/or opinion? Search on Google Groups for a thread called "4 the FAQ: Are Commercial DBMS Truly Relational" Most of the issues were dealt with in that thread. SignatureKenneth Downs Secure Data Software, Inc. (Ken)nneth@(Sec)ure(Dat)a(.com) >> The rating of 100 equates to full realisation of the RM. >> The rating of 50 equates to half realisation of the RM. >> The rating of 10 equates to 10% realisation of the RM. >> The rating of 0 equates to zero realisation of the RM ...[trim]... > Search on Google Groups for a thread called "4 the FAQ: Are Commercial > DBMS > Truly Relational" Most of the issues were dealt with in that thread. Some interesting responses, such as 'And, to tell the truth, I regard Oracle as "close enough for practical purposes".' In this case, the percentage might have been stated by this responder to be 90% (truly relational). Thanks for the reference. SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au >>> The rating of 100 equates to full realisation of the RM. >>> The rating of 50 equates to half realisation of the RM. [quoted text clipped - 14 lines] > responder to be 90% (truly relational). Thanks for the > reference. No problem. But what I liked about the responses was no so much the percentage idea, as the specific concepts. The percentage concept is then subjective. Depending on how important the divergent features are to you, that's how well the product conforms. SignatureKenneth Downs Secure Data Software, Inc. (Ken)nneth@(Sec)ure(Dat)a(.com) > No problem. But what I liked about the responses was no so much the > percentage idea, as the specific concepts. The percentage concept is then > subjective. Depending on how important the divergent features are to you, > that's how well the product conforms. The specific concepts are always interesting to consider and reflect upon however the concept that the SQL-DBMS "[abstract] machines" do not <INSERT SOME SUITABLE ADVERB HERE> (eg: truly, fully, etc) embody the relational model of data needs to be quantified by the proponents, seeing that, as they are so fond of saying, it is scientific. Although subjective consideration is always going to be happening, one might expect there *also* to be some general form of objective agreement, and this is what I wish to ascertain. Do relational theorists admit 50% "realisation", or do they admit 20% or do they admit 80%. Surely this is a question that may be answered by each (relational model) theorist independently, and so arrive at a consensus? SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au > How much now, as a percentage out of 100, are contributors > to this newsgroup prepared to estimate the "realisation" of the > relational model of data within the SQL-DBMS vendors > (ie: IBM - DB2, Oracle, MS - SQL Server) I think it's like asking "How much, as a percentage out of 100, was calculus "realised" in the 18th century by the current mathematicians", such as Lagrange and Euler. They surely made lot's of important discoveries, but still further important work was to come after the "cleaning up" of Analysis by the likes of Cauchy, Gauss and Riemann. For instance there was not even a proof for the binominal theorem when n is not an integer greater than zero until Gauss gave one. 1) So, from the perspective of the 18th century the answer would have maybe been close to 80, but we now know that they had barely scratched the surface! Likewise, since the RM is a mathematical theory, we don't really know where it will lead us in the future. However, it is important to try to keep the theory "clean" even though the practical consequences are not obvious at the time or might even cause some backsteps. The products (such as DB2, Oracle and SQLServer) are chained to their current user base and they don't really have a way out. They can only go towards increased complexity. That is not to say that they are not useful. But they are probably not in the position to take advantage of advances in the theory of the RM, or even realise it's current potential. Regards, Lauri Pietarinen 1) E.T. Bell, Men of Mathematics, p. 222 >>How much now, as a percentage out of 100, are contributors >>to this newsgroup prepared to estimate the "realisation" of the [quoted text clipped - 9 lines] > theorem when n is not an integer greater than zero until Gauss gave > one. 1) I think it was Riemann who put Calculus on solid ground with his idea of how to define continuity (at least that's what I remember from my Real Analysis class many years ago). I also think he went insane. And I don't blame him as he was a trail blazer. > So, from the perspective of the 18th century the answer would have > maybe been close to 80, but we now know that they had barely scratched [quoted text clipped - 15 lines] > > 1) E.T. Bell, Men of Mathematics, p. 222 > I think it was Riemann who put Calculus on solid ground with his idea of > how to define continuity (at least that's what I remember from my Real > Analysis class many years ago). I seem to remember that the epsilon-delta defintion was by Weierstrass, but that Bolzano had actually already defined it earlier. -- Jan Hidders >> I think it was Riemann who put Calculus on solid ground with his idea of >> how to define continuity (at least that's what I remember from my Real >> Analysis class many years ago). > > I seem to remember that the epsilon-delta defintion was by Weierstrass, > but that Bolzano had actually already defined it earlier. I seem to recall there is no "solid ground" in mathematics and that the formalisms of mathematics cannot lead to anything resembling "absolute truth". http://www.cs.auckland.ac.nz/CDMTCS/chaitin/cmu.html How does the RM get around Godel, Turing and Chaitin? SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au/software >> I think it was Riemann who put Calculus on solid ground with his idea >> of how to define continuity (at least that's what I remember from my [quoted text clipped - 4 lines] > > -- Jan Hidders http://en.wikipedia.org/wiki/Continuous_function >> How much now, as a percentage out of 100, are contributors >> to this newsgroup prepared to estimate the "realisation" of the [quoted text clipped - 9 lines] > theorem when n is not an integer greater than zero until Gauss gave > one. 1) Thanks for the journey of analogy through the history of mathematics. > So, from the perspective of the 18th century the answer would have > maybe been close to 80, but we now know that they had barely scratched > the surface! So is this an 80%, with reservations that this 80% will in 200 years time be viewed as 10%? > Likewise, since the RM is a mathematical theory, we don't really know > where it will lead us in the future. Is there only one Relational Model (ie: as espoused by Date) or are there variant interpretations thereof, such as there exist a variety of "intepretations" of quantum theory in physics? Also, it seems as though some people say it will evolve, such as yourself, yet others here assert it is an unchanging principle that does not change. At least we agree things evolve. > However, it is important to try > to keep the theory "clean" even though the practical consequences are > not obvious at the time or might even cause some backsteps. The > products (such as DB2, Oracle and SQLServer) are chained to their > current user base and they don't really have a way out. They can only > go towards increased complexity. They exist, it seems to me, as clear examples of incarnation of the RM, whether or not incomplete from the perspective of theory. There may be pathways towards simplicity. > That is not to say that they are not > useful. But they are probably not in the position to take advantage of [quoted text clipped - 5 lines] > > 1) E.T. Bell, Men of Mathematics, p. 222 What advances in the theory of the RM have there been since Codd's first paper, from the theory side? Are they numerous or singular? Thanks again for your contribution. SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au/software > What advances in the theory of the RM have there been since > Codd's first paper, from the theory side? Are they numerous > or singular? Do you mean stuff like normalisation theory, development of better relational operators, TABLE_DEE and TABLE_DUM, and the generalisation of relational operators to handle interval types and thus temporal data? SignatureJon > > I think it's like asking "How much, as a percentage out of 100, > > was calculus "realised" in the 18th century by the current [quoted text clipped - 11 lines] > So is this an 80%, with reservations that this 80% will in 200 > years time be viewed as 10%? Who knows? We will all be dead by then, to quote a famous American president. > > Likewise, since the RM is a mathematical theory, we don't really know > > where it will lead us in the future. > > Is there only one Relational Model (ie: as espoused by Date) > or are there variant interpretations thereof, such as there exist > a variety of "intepretations" of quantum theory in physics? Over time things will settle. It took a long time to find a good definition for x! (i.e. the factorial for non-integer numbers, e.g. (4.5)! ) but an elegant interpretation was found. As for the variant interpretations, some of them result from "fixing the theory around the practice" as with duplicate rows. I can't seriously see that as an interpretation. Others, such as allowing or not allowing nested relations, will - and have been - settled over time. Anyway, if we restrict the model not to include nested relations, we are not losing that much power in practice. Do you have other other interpretations in mind? > Also, it seems as though some people say it will evolve, > such as yourself, yet others here assert it is an unchanging > principle that does not change. At least we agree things > evolve. Surely the basic principles and conclusions of Newton and Leibniz still hold regarding Calculus inspite of the refinements made to it over time. An integer is still an integer albeit having been defined in various ways over the centuries. It's just that we will discover new ways to refine the definition and to build on it. So I believe that the very basic core will remain but it will be refined (and a refinement does not mean adding support for duplicate rows). And there will be more and more discoveries around this core, such as real and complex numbers in the case of refining the concept of a number. > > However, it is important to try > > to keep the theory "clean" even though the practical consequences are [quoted text clipped - 6 lines] > of the RM, whether or not incomplete from the perspective > of theory. There may be pathways towards simplicity. What incarnations do you have in mind? > What advances in the theory of the RM have there been since > Codd's first paper, from the theory side? Are they numerous > or singular? - Normalisation theory (strictly speaking not part of the RM) - Better understanding of transformation rules - Also lots of new theory regarding how to *implement* the RM, maybe not directly related to RM but at least inspired by it (optimisation, indexing strategies, locking). Some advances, such as sub tables, have turned out not to be very useful (they were inspired by the OO-model). I would not know how to answer your question with any great authority, but I hope you get the picture. To be frank, I am not totally comfortable with the idea of a group of people acting as the guardians and gate keepers of the RM, but I am much less happy of all the strange things proposed for data management, such as XML. "All models are wrong, but some are useful" So it is not really a question of if the RM is "God's answer to data management" but the question is "is it useful?" We don't know if Analysis is "correct" but somehow, mysteriously, it is useful for calculating many practical things. Logic, as a branch of mathematics, as turned out to be immensly useful. The RM is based on logic and hence can borrow from the results of 2500 years of work. regards, Lauri Pietarinen >> How much now, as a percentage out of 100, are contributors >> to this newsgroup prepared to estimate the "realisation" of the [quoted text clipped - 9 lines] > theorem when n is not an integer greater than zero until Gauss gave > one. 1) Thanks for the journey of analogy through the history of mathematics. > So, from the perspective of the 18th century the answer would have > maybe been close to 80, but we now know that they had barely scratched > the surface! So is this an 80%, with reservations that this 80% will in 200 years time be viewed as 10%? > Likewise, since the RM is a mathematical theory, we don't really know > where it will lead us in the future. Is there only one Relational Model (ie: as espoused by Date) or are there variant interpretations thereof, such as there exist a variety of "intepretations" of quantum theory in physics? Also, it seems as though some people say it will evolve, such as yourself, yet others here assert it is an unchanging principle that does not change. At least we agree things evolve. > However, it is important to try > to keep the theory "clean" even though the practical consequences are > not obvious at the time or might even cause some backsteps. The > products (such as DB2, Oracle and SQLServer) are chained to their > current user base and they don't really have a way out. They can only > go towards increased complexity. They exist, it seems to me, as clear examples of incarnation of the RM, whether or not incomplete from the perspective of theory. There may be pathways towards simplicity. > That is not to say that they are not > useful. But they are probably not in the position to take advantage of [quoted text clipped - 5 lines] > > 1) E.T. Bell, Men of Mathematics, p. 222 What advances in the theory of the RM have there been since Codd's first paper, from the theory side? Are they numerous or singular? Thanks again for your contribution. SignaturePete Brown IT Managers & Engineers Falls Creek Australia www.mountainman.com.au/software "mountain man" <hobbit@southern_seaweed.com.op> wrote in message news:... >> Dan wrote: >>> [quoted text clipped - 11 lines] > http://www.cs.auckland.ac.nz/CDMTCS/chaitin/cmu.html > How does the RM get around Godel, Turing and Chaitin? See separate thread. |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.