[ruby-core:43087] Re: feature request: marshallable proc's

On Mon, Mar 5, 2012 at 8:12 PM, Eric Hodel <[e-mail]> wrote:
> On Mar 5, 2012, at 8:51 AM, Roger Pack wrote:
>>> What should happen on the following code?
>>>
>>> f =3D open(something)
>>> a =3D background_process
>>> a.send {
>>> =A0f.puts 'hello'
>>> }
>>
>> Hmm. Either an except on the puts call, an exception on the 'send'
>> method
>
> What about:
>
> f =3D open(something)
> a =3D background_process
> a.send {
> =A0eval "f.puts 'hello'"
> }
>
>> or automagic Drb to call back to the element in the calling
>> program :)
>
> What about:
>
> f =3D open(something)
> a =3D data_store
> a.store {
> =A0f.puts 'hello'
> }
> exit
>
> # new process
>
> a =3D data_store
> a.run_stored
>

I believe people would expect the same behaviour regardless of whether
the proc was transmitted via Drb or saved to disk and later loaded
into a new process, which really only leaves the options of throwing
an exception when trying to marshal the proc or when trying to run the
puts method.

Since metaprogramming might make it pretty difficult to tell whether a
proc is doing anything illegal before marshalling it, I think throwing
an exception when it is loaded and re-run should be the way to go.

It's what Gemstone does, and they've used that stuff for years, so I
guess it has worked well for them.