stéphane ducasse wrote:
> Andreas just says that lukas is an idiot. This is more direct.

No, I said that the value of that change doesn't outweigh the value of 
the existing investments. And, I am now saying that I would appreciate 
it a *lot* if you weren't constantly putting words into my mouth. I'm 
quite capable of saying what I mean, thanks.

But to come back to the topic, if this were day 1 of the FFI I'd 
probably even go for the change in syntax. But as it stands I have to 
consider the investments of the people who are using the FFI. This is 
why I am the maintainer.

And as an advocate of my users I will tell you: Unless you provide some 
real tangible benefit for the users of the FFI, that change is not going 
to happen.

> Do not tell me that having a migration period would not work. 

I said something entirely different, namely that a migration only makes 
sense if a change were either desirable or necessary. The change that 
has been proposed is neither desirable (since its adds no value for the 
FFI users) nor necessary. Therefore no migration period is necessary 
since there is no good reason for a change to begin with.

> And do not tell us that the hacks introduced in the compiler are nice.

I haven't said that anywhere. However, regardless of their nicety, hacks 
are often useful if you deal with the reality outside of your personal 
view of what the world should be like. And in this case I wouldn't even 
consider it a hack - why do you call it that? The FFI has been present 
in the parser *far* longer than the pragma implementation. Given that 
this is the case you can hardly call it a "hack" - it is simply an 
extension to the parser added at (what at that time seemed to be) the 
right place.

I will admit that some of the code (in particular what Marcus' stumbled 
across) is a little questionable since it tries to avoid dealing with 
the dependency issues. And I have by now *twice* offered to discuss ways 
to deal with the problem, both times ignored.

It is certainly striking how you keep coming back to a no-value 
proposition and entirely avoid discussing what is an actual issue that 
should be addressed. That's what I mean by "ivory tower" behavior.

> What lukas was 
> proposing was a really nice way to integrate FFI with pragmas without 
> having these compiler hacks.

Sure. And it was a proposal that completely ignored any realities of the 
FFI. The FFI is there, it is used and people have invested in it. If you 
want to break all code anyone has ever written using the FFI I, both 
personally as well as a maintainer watching out for the interests of my 
users, want a reason for it. A Real Good Reason.

Cheers,
   - Andreas