Why no real RM-DBMS'es
Hi everybody, I'm Lauri, an SQL-addict since 32 years. Nice to be here!
I had the good fortune to start my DB-career by reading Chris Date's book "Introduction to DB2" and have tried to keep up with everything Chris has since written, which is quite challenging, since, I suspect, Chris can write faster that what I can read!
For a long time I have been wondering how SQL became what it did, and why there are no "real" relational DBMS'es available, even as prototypes. I was happy to find and read Hugh Darwen's article on this very topic, which does shed some light on the issue.
It still does not explain the lack of interest in the academic community on the subject. Of all the hundreds and thousends of IT-related research groups there seems to be more interest in chasing the lated fads. I can recall at the VLDB coference in Berlin in 2003 talking to a youg indian chap, who moved into DB research and found no activity in the "traditional" RM part, but now happily got a research paper on XML-indexing accepted in the conference.
At the 2001 VLDB in Vatican I managed to have a word with Jim Gray. I asked him what he thought of the TTM. I don't remeber exactly what he said but I remember his parting words, before being pulled away by his Microsoft handlers: "XML might, just might be *the* next data model". It makes you wonder... I see that Hugh has an article on that topic also.
Regards,
Lauri
There are some true relational DBMSs and related projects, listed at http://thethirdmanifesto.com/, but as you pointed out -- academia is very fad-driven. The RM was considered academically and the research community moved on, having essentially decided that if the problem involved relational databases, the solution was Datalog, the end. What's the next problem? Etc.
In the commercial world, SQL is good enough. No alternative, historical or modern, relationally "true" or otherwise, has yet demonstrated enough improvement to the workflow of the average SQL DBA, database designer, or application developer to drive any significant grassroots move away from SQL, and of course from a governance point of view, there are strong justifications to stick with convention and familiarity and keep using SQL.
Quote from Dave Voorhis on March 28, 2019, 7:30 amIn the commercial world, SQL is good enough. No alternative, historical or modern, relationally "true" or otherwise, has yet demonstrated enough improvement to the workflow of the average SQL DBA, database designer, or application developer to drive any significant grassroots move away from SQL, and of course from a governance point of view, there are strong justifications to stick with convention and familiarity and keep using SQL.
Indeed, working in the commercial world my self as a self empolyed database consultant, I understand the sticking power of SQL more than well. In some ways, it's a miracle that we *do* have SQL, we could still be chasing pointers.
But, one would think that, some academic group, somewhere, just by accident, just became curious as to the discrepancy between SQL and theory and tried to figure out what was going on. Maybe there is?
Is there any peer preview on TTM? Something? Is this the only group in the universe that is interested in this subject?
Lauri
Let me put it differently. Anyone interested in the subject ultimately ends up here, either as a lurker or in a smaller number of cases as an active contributor and in an even still smaller number of cases as an active implementer.
I don't know what you mean by "peer preview on TTM". Dave pointed you to the projects list. Projects have web sites. (Whose size is a linear function of the time available to write them, so that's a nice thing to know as a future reader.) There's more than enough for you to try out. If you want a pre"view", you ***are*** going to have to do the "viewing" part yourself.
Ah yes, there are quite a number of projects - I'll have to take a look at them! I ment "peer review", of course! Sorry about that. Let me ask that question in another way: "Is there any research activity going on around TTM, other than by D&D? Or maybe some alternative aproaches?"
Quote from LauriP on March 28, 2019, 8:15 am
But, one would think that, some academic group, somewhere, just by accident, just became curious as to the discrepancy between SQL and theory and tried to figure out what was going on. Maybe there is?
It's well-documented by Date & Darwen and others, so there's not enough mystery about why we've got SQL rather than something else to warrant real research effort. An hour or two of googling will answer any questions. Academic research groups generally have better -- and by "better", I mean "researching funding attracting" -- things to do.
Quote from LauriP on March 28, 2019, 8:15 am
Is there any peer preview on TTM? Something? Is this the only group in the universe that is interested in this subject?
Do you mean peer review?
There isn't a published "Journal of TTM" outside of this forum, which is discussion-based and does not use anonymous review. I've occasionally offered to create and sponsor the (new) "Journal of the Relational Model", which would be an open peer-reviewed e-journal, but assent has been lukewarm at best. I'm still keen to do so, if there's interest.
To my knowledge, this is the only group interested in the subject. I'm not aware of anyone working in the field who hasn't either been discussed here and has some peripheral connection to participants here, or is (or at least has been) a participant here.
Quote from LauriP on March 28, 2019, 9:57 amLet me ask that question in another way: "Is there any research activity going on around TTM, other than by D&D? Or maybe some alternative aproaches?"
Academia doesn't regard TTM as "academic work" so basically they'll take the stance that "there simply is nothing to review" (e.g. I recall a former ICDT chair saying he was "sympathetic" to the TTM "movement" but complained it was all way too light on the formal proofs side). Sometimes an academic takes the bother to put his criticisms in writing (google for criticisms by one Maurice Gittens and responses). Sometimes there are specific circumstances that make D&D's works attract more attention for a while. The introduction of the SQL:2011 standard put the Temporal Data stuff in the spolights (e.g. with LI discussion groups dedicated to the subject) for example. But in general the answer is "no".
Quote from LauriP on March 28, 2019, 9:57 amAh yes, there are quite a number of projects - I'll have to take a look at them! I ment "peer review", of course! Sorry about that. Let me ask that question in another way: "Is there any research activity going on around TTM, other than by D&D? Or maybe some alternative aproaches?"
Aside from rare and largely "out of band" papers, theoretical research on the relational model largely ended in the 1980's. There's still plenty of applied work -- which includes our various implementations. For example, I occasionally see references to Rel in MSc. dissertations and the like.
Quote from Erwin on March 28, 2019, 11:31 amQuote from LauriP on March 28, 2019, 9:57 amLet me ask that question in another way: "Is there any research activity going on around TTM, other than by D&D? Or maybe some alternative aproaches?"
Academia doesn't regard TTM as "academic work" so basically they'll take the stance that "there simply is nothing to review" (e.g. I recall a former ICDT chair saying he was "sympathetic" to the TTM "movement" but complained it was all way too light on the formal proofs side). Sometimes an academic takes the bother to put his criticisms in writing (google for criticisms by one Maurice Gittens and responses). Sometimes there are specific circumstances that make D&D's works attract more attention for a while. The introduction of the SQL:2011 standard put the Temporal Data stuff in the spolights (e.g. with LI discussion groups dedicated to the subject) for example. But in general the answer is "no".
TTM is academic work -- applied work, largely -- but academia is heavily driven by fashion and perspective. For every researcher whose perspective leans toward "formal proofs", there'll be another focused heavily on practical implementations. Indeed, I recall one correspondent who categorically rejected TTM because the presented Tutorial D grammar was incomplete.
It's mainly academic fashion that keeps TTM sidelined. If you want to be funded, today your research needs to focus on machine learning, predictive analytics, Internet of Things, or high performance computing. Next year, it will be something different. Herds of researchers lurch from topic to topic as popularity shifts. They're typically not even interested in the topics they are researching; they're just interested in "doing research" and publishing papers. The topic doesn't matter.
For publication purposes, TTM could probably ride on the -- also more applied than theoretical -- NoSQL movement, where there's certainly some current academic publication activity, but I don't get the impression most of us are particularly interested in formal publication. I'm not, and I work in academia. At least, for now -- I'm retiring this summer, at least from academia and maybe from employment in the usual sense, though my work on Rel and related projects will hopefully increase as a result. That's what really interests me: making software.
Quote from Dave Voorhis on March 28, 2019, 12:17 pmQuote from Erwin on March 28, 2019, 11:31 amQuote from LauriP on March 28, 2019, 9:57 amLet me ask that question in another way: "Is there any research activity going on around TTM, other than by D&D? Or maybe some alternative aproaches?"
Academia doesn't regard TTM as "academic work" so basically they'll take the stance that "there simply is nothing to review" (e.g. I recall a former ICDT chair saying he was "sympathetic" to the TTM "movement" but complained it was all way too light on the formal proofs side). Sometimes an academic takes the bother to put his criticisms in writing (google for criticisms by one Maurice Gittens and responses). Sometimes there are specific circumstances that make D&D's works attract more attention for a while. The introduction of the SQL:2011 standard put the Temporal Data stuff in the spolights (e.g. with LI discussion groups dedicated to the subject) for example. But in general the answer is "no".
TTM is academic work -- applied work, largely -- but academia is heavily driven by fashion and perspective. For every researcher whose perspective leans toward "formal proofs", there'll be another focused heavily on practical implementations. Indeed, I recall one correspondent who categorically rejected TTM because the presented Tutorial D grammar was incomplete.
It's mainly academic fashion that keeps TTM sidelined. If you want to be funded, today your research needs to focus on machine learning, predictive analytics, Internet of Things, or high performance computing. Next year, it will be something different. Herds of researchers lurch from topic to topic as popularity shifts. They're typically not even interested in the topics they are researching; they're just interested in "doing research" and publishing papers. The topic doesn't matter.
This fad-orientation is really strange and seems so counter productive, but I saw it myself first handed. And OK, if 95% is chasing the next thing (XML indexing in early 2000 and now, whatever), what are the rest doing? Or would that be 100% just following the crowd like lemmings?