snax

The CRUD debate continues, here and here .

semantic resources

Regarding his first point, I agree. I didn’t mean to suggest that the implemented representation wasn’t separate from the semantic content of a resource. Certainly the server doesn’t save exactly what the resource “is”. However I think it is too far to say that anything named by a URI is a resource. Semantic continuity can easily be broken. If a domain name changes hands and is repurposed, it is not the same resource anymore merely because it is still the same URI (to take an extreme example).

In CRUD, the server does not perform semantic transformations on the resource.

multiple models

“You just need a model that’s the appropriate projection of the multiple.”

Emphasis on “a”. It’s not multiple models anymore if you had to make a separate model to represent it. Also, you’ve added an avoidable layer of complexity, and CRUD is about keeping it simple.

command pattern

The command pattern has many uses, but it’s not for everything. Think what would happen if in Ruby the only methods allowed on objects were CRUD, and possibly “execute”. Furthermore, I don’t think that spinning out all possible parameter configurations into logical URIs really counts as CRUD, but I have a feeling I am going to be disagreed with. But do you want:

ReverseString.new(:string => "string") 
ReverseString.execute
ReverseString.read

or

"string".reverse

And if image crop is not a good example, take protein similarity search instead—something that can’t possibly be offloaded to the client.

target domain

Again, I think that most of this debate is due to the limited target domain of a lot of Rails applications. In many consumer applications, CRUD is all you need. But that does not mean that CRUD is all you need anywhere.

if all_you_have.include? hammer and all_you_have.length == 1
  everything.each { |thing| thing = Nail.new }
end

August 20, 2007

2 comments

Comments are closed.