[3101] in java-interest
Re: dynamic method invocation -- any more discussion?
daemon@ATHENA.MIT.EDU (Pat Niemeyer)
Tue Oct 31 01:43:39 1995
Date: Mon, 30 Oct 1995 23:17:22 -0600 (CST)
From: Pat Niemeyer <pat@icon-stl.net>
To: "Steve V. Yalovitser" <root@cyberpunk.com>
cc: java-interest@java.sun.com, Chris Cuilla <ccuilla@m-space.com>
In-Reply-To: <951030190959.0.adc949@sensenet.cyberpunk.com>
On Mon, 30 Oct 1995, Steve V. Yalovitser wrote:
> Dynamic method invocation was relatively simple to hack up for the simple
> case. There was an old mail by Tom Ball or Arthur which gave a blueprint. It isnt
> really correct but should give a start:
[ something deleted]
> Mike Responds...
>
> >Sounds cool however, Java is not a dynamic language in the sense that
> >Smalltalk is a dynamic language. This works it self back to the request for
> >perform: capabilities ala Smalltalk. The reason for no perform: (as it was
> >explained to me) is that it opens a big security hole. For example, a Nasty
> >Old Java App (NOJA) could evoke the method "MySystem.sendUglyMailToMyBoss"
> if
> >a class with such an interface exists. I guess that the NOJA could scan the
> >class which it finds on a given workstation and try things at random causing
> >all sorts of havoc.
I don't think this is the rationale behind no method pointers. I've been
trying to dig through my mail and find the one authoritative comment on this
that I remember, but I seem to have lost it...
I thought that the response was that the Java people thought long and hard
about method pointers and ultimately there were issues involving "queuing"
of method invocations or something... There was even a mention of dredging
up some old mail on the subject, but it never surfaced.
I don't think that I've heard any real security issues involving dynamic
method selection. Run time checks are there anyway...
Anyone have that mail or did I just dream it?
Pat
-
This message was sent to the java-interest mailing list
Info: send 'help' to java-interest-request@java.sun.com
