Re: Q: type ... is new String

[Georg Bauhaus <rm.dash-bauhaus@futureapps.de>]

On 05.06.12 11:06, Dmitry A. Kazakov wrote:

>> (All I see is subsequences of 2#bbbb_bbbb#. For the AI part,
>> I am told to produce the most likely information that the
>> originator might have intended to send.)
> 
> Bad design. Don't do that.

How is producing the most likely information from a piece
of data bad design? And who am I to say "I don't do that"?


> Do not conflate
> well-defined functionality, e.g. "send over socket", "render calendar page
> on the screen" with ill-defined stuff, like what is going on in someone's
> head.

What is going on in someone's head is precisely the information
I am looking for. It so happens that this includes resource
identifiers, and here, too, I need to guess the right one, because
the transport layer conveys mistakes, omissions, and
quadruped-on-the-keyboard-effects correctly.

And, again, I must rely on guesses in order to construct objects
from input, and will reconstruct them again, and again.

Is this about a trivially correct ADT again, to be fleshed out?

I can make a sufficiently simple type for URIs, signifying nothing,
and I can map any of those resource identifiers to objects of the type.
I do not even need a behavioral description of the type; it is
exactly that of strings (including a hashing function in the
language used).
So all the separate type does is that it forms a subset of strings.
Good.

That seems pretty close to Spolsky's safe/unsafe naming convention,
but better, if refactoring is needed, because a type is better
than an Emacs macro (to which he is referring somewhere).

> For broken design it is always too late. Fix the design.

The design as a whole is not in our hands, I should think. The design
is as is, dictated by a combination of current input and time available.

> How an incomplete definition is job done?

If
- foreign company's management says "Job done!" and
- foreign company works profitably,
= then an incomplete definition means job done.

The resulting "designed system" is not ideal, from my technical
point of view. But they need not care.


> Why should I care about description of naming schemes? I need a set of
> types describing file name valid in some specified environment.

Whenever there is a sufficient specification of an environment, fine.

Just saying that web documents are an example of when a specification
might be insufficient, if it exists at all.