On Sun, 2011-03-06 at 15:34 +0100, Chris Young wrote:
> On Sun, 06 Mar 2011 14:12:11 +0000, John-Mark Bell wrote:
> 
> > I'm really not convinced that using the plugin hooks like this is the
> > right way of solving the problem you appear to be trying to solve.
> > Particularly, treating plugins as valid image content types likely opens
> > a large can of worms (not least, as plugin support was disabled because
> > it resulted in many crashes)
> 
> Without treating them as images none of the plugin functions are
> called for objects embedded on pages... although actually, I see that
> is my mistake as of course plugins are usually embedded differently.

Indeed.

> > Could you explain what it is you're actually trying to do here?
> > 
> > If, as I suspect, it's to be able to dynamically support additional MIME
> > types, then the better solution is to make the content type dispatcher
> > able to accept runtime registration of MIME types. This is something
> > that is planned, but won't happen until after 2.7 as it requires major
> > surgery to the core.
> 
> Yes, that's pretty much it.  I was focussing on images initially in a
> bid to remove some of the larger static libs from that build, as it
> tends to get used on low memory classic machines.

Ok. Sounds like we need to bump the priority of dynamic content handler
registration, then.

> plugin_handleable and treating plugins as images works for this,
> supporting embedded audio might need plugin_open or the mouse action
> function.

Handling non-visual contents is a whole other can of worms. :)

> The ordering of functions and some of the parameters were wrong for
> RISC OS, but I suspect these are just things that hadn't been updated
> along with the core rather than anything to do with the crashing.

Yeah; that code simply hasn't been built for ages, hence why it gets
broken every now and again.

> I'll remove plugins from image_types to stop it causing more trouble.

Ta.


J.


Reply via email to