Re: [Sbcl-devel] some questions regarding make-thread

[Attila L. <att...@gm...>]

> I'm about introducing an sb-thread:*default-special-bindings*. Seems
> like other Lisps implement that as an alist between variable names, and
> forms evaluated in the new thread. They also seem to make the variable's
> default value to be the bindings behind WITH-STANDARD-IO-SYNTAX.


i propose to use closures instead of eval.


pro arguments:

 - funcall'ing a closure is much faster than eval

 - when applicable, a closure is a more flexible way of delaying
computation than eval

 - slot initforms are a very similar issue, and they use closures


cons arguments i don't agree with:

 - eval is easier to debug -- if so then don't use closures in
general, or fix the debugger (which is not needed, slime works good
enough in this sense)

 - it's harder to type the extra lambda which is not needed 99% of the
time -- imho it's better to have a flexible atom with a thin macro
hiding the details (if needed at all), than having a slow and less
flexible atom slightly easier to use directly. and this feature is not
something you'll type several times a day...

 - everyone does it the eval way -- this is not a technical argument.
and cross platform libraries can hide away this difference easily (if
they forbid initforms to use the lexical environment)

-- 
 attila