Re: Avoiding explicit addDataType calls for PostGIS

[Sorry for the duplicate message, Oliver, my fist mail was
unintentionally sent privately to you]

Hi, Oliver,

On Thu, 07 Oct 2004 23:31:25 +1300
Oliver Jowett <oliver@opencloud.com> wrote:

> > Is the opposite problem possible? I think of the driver class loader be
> > able to reach the driver extension, but not the user code class loader.
>
> That seems unlikely; it'd only happen if you had user code that did not
> contain any references to the extension classes. If there are no
> references to the extensions, how does the user code use the returned
> objects? Reflection, or a common superclass shared by both classloaders,
> seem like the only options.

It is also possible that the classes were known at compile time, and are
used read-only at run-time. That means that the user does not create
instances itsself, but only gets them from the db.

But I assume this are academic examples.

> It's also debatable whether the driver should even allow user code to
> cause loading and instantiation of classes that the user code would not
> otherwise be able to access..

That's a good question.

So, from the security point, we should scan the ressources for possible
classes (not eliminating collisions), and then filter all classes the
user gives by name (via addDataType, URL or whatever) wether they are
allowed (they exist in the ressources file).

Markus

--
markus schaber | dipl. informatiker
logi-track ag | rennweg 14-16 | ch 8001 zürich
phone +41-43-888 62 52 | fax +41-43-888 62 53
mailto:schabios@logi-track.com | www.logi-track.com