In comp.os.linux.advocacy, [H]omer
<spam@xxxxxxx>
wrote
on Wed, 31 Oct 2007 14:36:32 +0000
<iggmv4-rrp.ln1@xxxxxxxxxx>:
> Verily I say unto thee, that The Ghost In The Machine spake thusly:
>> In comp.os.linux.advocacy, Keith Windsor <keith@xxxxxxxxxxx> wrote on
>> Tue, 30 Oct 2007 15:39:10 -0400
>> <47277c7c$0$26469$88260bb3@xxxxxxxxxxxxxxxxx>:
>
>>> It would be better to simply send the 'object' across and
>>> automatically remap the address space of the object across the two
>>> scripts without having to copy the data.
>>
>> Gosh, how "innovative"! The Amiga was doing that, albeit not at the
>> shell level, back in the mid-80's! :-)
>
> AmgiaDOS had access to shared objects through both Workbench and the
> console via ARexx, provided that the supporting application had the
> necessary Arexx port and allowed such object access internally (as many
> did). Of course there were other mechanisms too, such as the
> multifarious handlers in SYS:L.
Actually, I was thinking the Message/Port stuff at a lower
level, but yes, Amiga was very good at passing objects
in system. Outside of system was a little harder, and
I don't think AmigaOS quite worked that out.
But then, this was before C++ and Java, after all.
>
>>> This would be especially useful if the second script that accepts
>>> this object as an input parameter did very little with the object
>>> then say... see if the collection is empty or not.
>
> There should be no restrictions on what can or cannot be done with
> shared resources, provided the correct locking mechanisms are used,
> and the right security contexts.
>
>> In which case one might simply pass a count instead.
>
> Indeed.
>
> Both pipes and shm are useful in different circumstances, in particular
> shm is of little use WRT network resources, virtualisation, sandboxing,
> printing, or frankly anything other than in-place operations that
> typically preclude backup objects.
>
I can get real Smalltalkish, admittedly; what is an object
anyway? At best, it's something that can receive a message
and contains data (and in a pinch one can skip the data,
resulting in, among other things, proxies -- of course the
proxy has to be able to find its destination).
At worst, it's a phantom concern; it all reduces to
machine language in the end. (Which machine, of course,
is negotiable.)
Java/EJB, for example, is an interesting implementation of
proxies autogenerated by the client side (the communication
uses a proprietary binary RMI, which IINM is a variant of
the .class file format.)
--
#191, ewill3@xxxxxxxxxxxxx
Q: "Why is my computer doing that?"
A: "Don't do that and you'll be fine."
--
Posted via a free Usenet account from http://www.teranews.com
|
|
