The database curmudgeon speaks again

Hi all,

Normally I would have posted this in the General Database Discussion forum, but the discussion ranges beyond data management to the IT industry in general.

Everybody's favorite database curmudgeon is quoted here in a good overall discussion on the state of the data management, and the computer/programming industry in general: http://www.wilshireconferences.com/...iews/pascal.htm

I tell ya, in spite of how Fabian Pascal rubs people the wrong way, what he says is so true. Read near the bottom:
"...most of the time and effort goes into migrating from one fad to another, in mapping from one model to another, and in trying to make all the disparate technologies and acronyms work together. It’s a colossal waste of resources..."

That, my friends is the computer industry in a nutshell. We are constantly chasing our tails, and honestly, is software all that much better now than it used to be 20 years ago? No, because now we have 421 ways of doing the same thing, instead of 5. So, we spend half the time just grinding our mental gears trying to decide how this will talk to that, and what "model" to use, and whether it will be obsolete in a year, etc...

Maybe its time to go back to the basics and try to get them right this time .

Interesting indeed....

But guilty as charged, am I. To be honest, I feel as though I've fallen into the trap of "needing" the "latest and greatest <whatever>" and didn't even realize it. After reading the review, I read some of the things on Database Debunkings (the InnoDB interchange was particularly interesting - even I got the imression the InnoDB developer was a bit out of his league and didn't know what he was talking about, and I'm a total n00b on the subject) and gathered some info on Codd and relational algebra. It's amazing how quickly higher level things start to make sense when you start learning the fundamentals....

The one thing I can attest to firsthand, unfortunately, is Fabian's statement about academia becoming a glorified training department for vendors. Schools are offering programs to "get you ready for the next revolution" in the computer world, but by the time people graduate, their skills have been obsoleted by the nex, next revolution!

I've looked at XML Databases, and I'm studying XML right now. I'm going to be honest with myself and say this: "I do not see a friggin' point to XML at all as it relates to technology professionals". I tried to convince myself for awhile that "Oh yea! This makes sense!". But, to me, it doesn't. The structure that XML provides is, IMHO, convoluted and too difficult to express for real data management. It makes a nice language for expressing the structure of a document so you can print it on 87 different mediums without reformatting it 87 times, but for crying out loud... it's useless for actually MANAGING the data. We should be teaching it to publishers and secretaries... not IT people and programmers...

This guy interests me though, so I'm going to go read some more of his articles, and maybe even get one of his books for the time being.

That's funny you say that about XML, that has been in the back of my head for awhile. Everytime I look at it I keep thinking ok, what can I use this for, and everytime it's just a big nothing.

A few days later, I think I must be missing something....

I find myself thinking the same exact thing you two do about XML.

This is exactly what Fabian Pascal argues. See the following:

http://www.dbazine.com/pascal8.html
http://www.dbazine.com/pascal9.html

And here, C.J. Date argues that a truly relational DBMS should be able to support an XML data type:

http://www.dbdebunk.com/lauri1.htm

(Of course, PostgreSQL is an example of a DBMS with extensible types )

But yes, in general, XML is really just a new twist on an already-discarded data management concept: hierarchical databases. The whole problem with hierarchies is that you find yourself spending more time worrying about the path to your data than the data itself. We had a good discussion on this earlier.

But again, back to my larger philosophical point: the IT/Development world is becoming like ancient Rome or Greece, where there was a statue erected to every god you might think up, just in the hope of placating all the gods, and not missing any.

Nowadays, you can choose between hundreds of different programming languages, thousands of formats and protocols, and at least 30 major development models, and who knows how many approaches to data management, and 50 development frameworks, for each branch of computing you might work in. If we give them all "equal time", we will never have time to properly evaluate what is worthwile, so we tend to just pick a combination, based on ill-informed opinions, popularity, and marketing releases, and set about defending it as if our lives depended on it. We spend more time learning how other people think we should do things than in actually doing them.

This is why I wish sometimes we could go back to the basics, focus on learning our craft, instead of learning all the layers of junk that have been placed over everything. I even think PHP is becoming bloated. There is a function for everything; in fact there are multiple functions that do almost the same thing. And honestly, most of the more specialized functions are unnecessary, because we can accomplish them with a line or two of "regular" PHP code, and have a much more efficient runtime.

The problem is, too many people stop learning too soon, and simply assume that they have come up with something new, and start trying to sell it to everyone else. Talk about muddying the waters...

In contrast to all this, I think about Donald Knuth, who spent half a lifetime writing 3 thick volumes about how much you could accomplish with a 50-instruction assembly on a computer with only a few registers, and no OS (the mythical MIX computer). It's time to bring some of this art back into computing.

I sometimes think the same thing. Have you looked at LXP?

If it wasn't tied to linux and ran natively under FreeBSD I'd be tempted to try it.

its not open source!

The language I have been thinking about trying is TCL. And, there are Apache modules for it . There was a discussion on Slashdot and OSnews about it recently, and I had to admit it looked interesting. Also, many developers say the C source to TCL is some of the cleanest, manageable C they have ever seen.

Don't get me wrong, I still like PHP. My complaint about PHP is not really its size, but that so many of the specialized functions are included in one big jumbled-up namespace. At least with Perl, most of the specialized stuff is kept in modules. But, it does look like there are people working hard to fix certain aspects of PHP.

(But download mod_tcl and mod_dtcl and you will see the perspective I am talking about: these are tiny compared to PHP)

It's nice to have extensions to a language and be able to benefit from other people's complex solutions to implementation problems, but sometimes it feels like every time someone writes a line of code anymore, they feel obligated to "share it" with the rest of the world. Coding is becoming less and less an exercise in problem solving and more and more a "seek and paste" operation on Google.


I've been reading up through Pascal's old articles from the start (or, at least the start of the ones I could find), so I'll eventually get to them Sometimes I think he's just a tad too critical of things, however. For example, PostgreSQL works just fine and does what I need it to do. Therefore, for what I'm working on, Postgres is a good system whether it's a TRDBMS or not. Otherwise, I tend to like what he says and agree (almost) whole-heartedly.