Just A Summary : Martin Fowler's big mouthful

Comment on Martin Fowler's big mouthful by Daniel Berger

Sat, 19 Jan 2008 02:09:22 -0600

I’d be curious to see how the Jifty example was implemented. Looks pretty good, actually.

Comment on Martin Fowler's big mouthful by she

Sat, 19 Jan 2008 05:48:31 -0600

hey piers, hope you dont mind a dissenting opinion about one thing (whereas I think the rest of the article is nice)

It is about the word “pidgin”. I think the word is not good to address “internal DSLs”. The whole DSL area seems to confuse people. I read someone else claiming that BLAST is a DSL (but BLAST is just a simple, “stupid” dataset!)

So the term DSL is already inflated due to various reasons (i myself prefer to define DSL as something from humans, for humans, and thus nicely readable but capable or “programmed” to do something useful), and now instead of referring as “internal DSL” someone is using “pidgin”. Now I dont at all think you should change your way to label this ;) but I want to present my point of view, that this adds to confusion in general about DSL and DSL usage.

Anyway… I hope it won’t become like Monads in Haskell in the long run ;)

Comment on Martin Fowler's big mouthful by Mark Snyder

Sat, 19 Jan 2008 11:24:55 -0600

Hmm, odd nomenclature… most of the literature I have encountered has settled on “embedded DSLs.”

This technique of building an AST under the covers is used quite often in the Haskell community. In Haskell, monads and phantom types are the “corners” of the language that makes this possible. And both are mechanisms which – to agree with you – are very specific to Haskell. Take a look at “Domain Specific Embedded Compilers” by Daan Leijen and Erik Meijer if you are interested.

As for Martin Fowler, I appreciated some of his earlier writings on UML and Object Oriented Architecture, but I am increasingly getting the feeling that he is out of his depth when discussing programming language issues.

Comment on Martin Fowler's big mouthful by Piers Cawley

Sat, 19 Jan 2008 20:48:54 -0600

She said:

I read someone else claiming that BLAST is a DSL (but BLAST is just a simple, “stupid” dataset!)

Um… the person who described BLAST as a DSL was me. And I still stand by that. Just because a language isn’t (usually) executable doesn’t disqualify it as a language.

As for why ‘pidgin’, it’s precisely to get away from the over inflation of the term DSL and an attempt to unpack the way that these things are an amalgam of the grammar of a host and vocabulary from the domain and the host language. I blogged about it back in August.

Comment on Martin Fowler's big mouthful by Was it a cat i saW

Sun, 20 Jan 2008 01:16:16 -0600

I agree with Mark: people are now using “embedded DSLs”. Why pollute the namespace with yet one more term that must be explained every time you use it?

Comment on Martin Fowler's big mouthful by Piers Cawley

Sun, 20 Jan 2008 03:45:32 -0600

Why use pidgin? Precisely because it doesn’t muddy the water around a disputed term like DSL – which is becoming so overused in certain circles as to become almost meaningless.

If you don’t want to use the term, you’re welcome not to, but I shall continue to use it, if only because, once it’s been explained, it’s shorter to type than ‘embedded DSL’ and the temptation to write ‘DSL’ where I mean ‘embedded DSL’ won’t be there at all.

Comment on Martin Fowler's big mouthful by Daniel Berger

Sun, 20 Jan 2008 07:23:53 -0600

I vote for DSI – Domain Specific Interface. As chromatic often says, it’s just an interface. But it’s, you know, domain specific. And more descriptive than ‘pidgin’. Plus, you still get a nice, short acronym. :)

Comment on Martin Fowler's big mouthful by Piers Cawley

Sun, 20 Jan 2008 15:22:10 -0600

Except that sometimes that sells what’s being done short. Admittedly, something like ActiveRecord’s family of class methods isn’t much more than a Domain Specific Interface (and a pretty good one at that), but pidgins like the one used in Jifty’s Schema definitions end up building an Abstract Syntax Tree (or something very like it) which is interpreted by different tree walking Interpreters depending on whether it’s being used to build the database, generate accessor methods or automagically build HTML forms and that’s language technology that is. The tree may be being built in a somewhat unorthodox fashion, but the backend(s) have all the trappings of a language.