Interesting! I grew up mostly in the Midwest and South, and unless you were doing science (of any type) there was no mention of "metric" things. "2 Liter" became popular because that was a convenient bottle size of soda (yet another regional term) to buy but that was about it.

There are certainly "schools" of programming but I don't think in the sense of a change in the language itself. The main reason for this is that all languages used in mainstream development are built around more or less formal specifications of the language. This usually means that some entity (standards body, company) owns the specification and can declare an implementation of the specification non-compliant. Witness the recent flap between Sun Microsystems and Microsoft over the Java Language. Sun owns the specification of the Java language and licenses the use of the specification based on compliance. The flap resulted when Microsoft attempted to move slightly (well, actually quite a bit) away from the specification to suit their own purposes. Now I can use the language to create "tools" that make doing certain things easier. These tools might be considered dialects in the sense that they are peculiar to my own development experience and are not in widespread use. If I want to change the behaviour of a language, I would have to approach the owner of the specification and request a change, but if I'm the only one asking for the change, it's not likely to happen. This is true for most "modern" languages.

It has been true in the past that variants of languages have become full-fledged languages in their own right. There were a family of languages originating (I think, must check) from Snobol that were effectively dialects of that language. They all had interesting names such Spitbol, Slobol, etc. The language Cobol is a distant relative of these languages.

There are different schools of programming centered around the types of constructs a language supports. The principal ones are procedural languages (such as C, C++, Java), declarative languages (IDL, SQL), and functional languages (Haskell, ML, Scheme). Of these three, the one that is most different is the set of declarative languages. The other two have in common that they provide language tokens that support operation sequencing, looping, sub procedures, and conditional sequencing and differ mainly in how they treat assignment of value to entities not in the token set.

Your scenario is probably more true than not. But the real comeuppance of the "cowboy" programmer group is that they probably make use of obscure details of the language to accomplish the same task in less code and maybe more efficiently. I can in fact write better code in C++ than some programmers because when I need speed I resort to calling down into the native assembly (machine specific) code to directly manipulate the machine, but not everyone can do this. But it also raises issues that are beyond use of the language. One goal of writing commercial software is that it must be maintainable. So if I use obscure language details or assembly code, I may make the software less maintainable and possibly unsupportable if future requirements mean changing to different languages (as often happens). This is important because frequently the people who maintain the software in the future may not be experienced programmers but interns or co-ops.

So I've been rambling, but the gist is that computer languages are more fixed because of the reliance upon standards/specifications. The dialect concept probably arise because of the use of language features that are not in common usage.

I've found that in just about every company I've worked for, the discussions are vast intellectual wastelands since most of the conversations revolve around sports, beer, or the stock option value.