在 2006/7/12 上午 1:12 時,Allison Randal via RT 寫到:

Audrey Tang wrote:

That is a sane argument, which is why I think punt-and-see has some merit:
as soon as there is a workaround forced to be expressed at :immediate
level, we can evaluate it and see if it's better handled declaratively.

Excellent. (Er, though you know that .loadlib isn't really any more
declarative than :immediate, right? It's a compile-time directive. That
is, it performs an imperative action at compile-time, just like the
equivalent 'loadlib' in :immediate.)

It is declarative because it only has one surface form, and dependency
analysis programs can easily parse that.  Compare to :immediate, which
is essentially impossible to grok correctly without using error-prone
heuristics.

So both are imperative in its effect; ".loadlib" is declarative and makes
toolmaker's life easier, because if I want to extract this information,
I'm not required to run unbounded arbitrary code.

Can we agree on the punt-and-see, neither-encourage-nor-deprecate
resolution for now, then?

Pretty much. Though I'd phrase it as :immediate will not be deprecated.
And we wholeheartedly encourage people to use :immediate wherever it
solves their problem.

That is fine. I meant "we don't encourage people to use :immediate when a
simple .loadlib would do -- we encourage them to use .loadlib instead."

Glad to see speedy consensus.  Thanks a lot for your time. :)

Thanks,
Audrey

Attachment: PGP.sig
Description: This is a digitally signed message part

Reply via email to