[Freeciv-Dev] Re: [PATCH] Names in rulesets. (PR#559)

[Erik Sigra <freeciv@xxxxxxx>]

Gaute B Strokkenes wrote:
> 
> Erik Sigra <freeciv@xxxxxxx> writes:
> 
> > Gaute B Strokkenes wrote:
> >
> > > I think a fatal error is a bit harsh, though.  And I think we
> > > ought to allow case variations.
> >
> > I don't think it is too harsh considering it can be easily fixed by
> > the ruleset creator.
> 
> So she fixes the error, and then five minutes later, when she gets
> around to running civserver again, civserver dies with another error.
> Repeat ad nauseum.  I think it's much easier on the creator to go on
> loading the rules and list as many errors as possible before
> (possibly) dying.  This is why gcc and friends don't die when they
> encounter the first error in file.  It would be a real pain to use
> them if they did.  (Off course civserver isn't gcc, but...)

OK.

> > In princlple I like things that force correctness.
> 
> Really?  Did you ever hear about "bondage and discipline" languages?

I had a 5-week programming course in the spring. We used Ada95. That is the 
best and friendliest language I have ever tested. And it is the only language I 
have ever recieved academic training in.

A programmer shouldn't have to waste time on experimenting with different 
formatting styles. The formatting style should be given so she can concentrate 
on the fun part, the actual coding.

> In principle, I think that software should be robust and not die when
> faced with errors that it can easily recover from.
> 
> > What would case variations be good for?
> 
> Why not?  So that you can write "male" instead of "Male", for
> instance.  There are a number of places where case sensitivity _is_
> appropriate, mostly because it would confuse gettext(), but I don't
> think this is one of them.

If we for some reason think that "male" is better than "Male" we should use it 
everywhere and not let chaos take over. Allowing random case increases entropy. 
Then the comression will be less effective. I made a little test: I took the 
first 10 nations (american to brazilian) and copied them to two directories. 
Let's call them nation/ and nATioN/. In nATioN, I randomized the case of the 
letters in the leader sexes. Then I tared the files in each directory. Both 
tar-files had the same size of course. Then I bzipped the tar files. 
nation/nations.tar.bz2 became 3966B large. nATioN/nATioNs.tar.bz2 became 4042B 
large. So I lost 76B only because of this very little change. The difference 
might become even bigger if it was made truly random. 76B is not much but it 
would add up if we allowed sloppiness in more places in the distribution. 
(Besides, most people think entropy is ugly.)

The increased size might be a little disadvantage but the advantage of being 
able to write "male" instead of "Male" in some places seems even litteler. That 
implies the conclution that case variations are no good.