Re: OLE Discussion (was Re: Opendoc)

[Stefan Nobis <snobis usa net>]

Miguel de Icaza <miguel@nuclecu.unam.mx> writes:

> > loaded. By the way: The show-methods for all objects (and the objects
> > themselves) have to be loaded at startup (and surely OLE does this
> 
> Nope.  As I mentioned, OLE2 delays the component loading until it is
> absolutely necessary:  if only the display of the embedded is
> requested and an available cached image is available, that will be
> used until the user actually activates the object.  Cool stuff.

Hmmm... in OpenDoc most read-only components, those which are
responsible for drawing the object, are less than 100KB (mostly only
some dozens KB). Also cool stuff. ;)

> > And loading everything at startup hurts much less in OpenDoc than in
> > OLE, cause OpenDoc classes are lean, OLE classes are very big
> > monolithic applications.
> 
> Oh?  Can you comment more on this issue?  I want to know what I should
> be looking for to avoid screwing us.

What do you mean? I played with a testcase from IBM for OpenDoc and
nearly all implemented classes were <100KB but i donīt remember the
amount of the OpenDoc runtime.

Why are OpenDoc classes so lean? Cause they only do what they should
do - just like the UNIX philosophie. A text-class only handles text
and no graphic and no tables and the like. A container-class is
responsible for aligning objects on a page (that class is for layout,
the DTP-functionality). Maybe there are classes for the date, clock,
tables, graphics and the like. Also you have to kinds of classes: The
edit-classes to really change an object and the viewer-classes, which
only shows you the content of an object - this way there are no
problems with commercial use (give the viewers away for no cost and
make your money with the edit-classes) and you have the advantage of
small and fast classes for display-only purposes.

Next week i will look for more detailed information on OpenDoc.

-- 
Until the next mail...,
Stefan.