Squeak Team Plans...
DanI at wdi.disney.com
Wed May 27 20:44:14 UTC 1998
>Any chance those internal enhancements include the ability to make the kernel
>reentrant and embedable?
Yes and no.
We are planning enhancements for the VM in the course of the summer. It's our hope to work with Andreas to clean up a bunch of things, and possibly fold in functionality similar to the Squeak API work, and maybe go to a plug-in architecture. Both of these will depend on finding some approach that doesn't seriously compromise simplicity, kernel size, or portability.
While we're at it, a pass at reentrancy would not seem inappropriate.
Along the lines of embedding, we are interested in ROMable and derivative (incremental) images. The ROOTS table in Squeak's object memory makes derivative images particularly simple, and one can certainly take an easy first step toward a ROM image by running the SystemTracer and telling it to put out all the CompiledMethods (read-only), MethodDictionaries and Classes (almost always read-only) at the beginning of the image.
So that is a sort of yes...
However, while these are all fascinating pursuits, none of these are as high priority right now for the Squeak team as making the top layer of Squeak utterly simple and powerfully general. I think we have a good record of moving the VM gradually forward, but I can't really say that we will have done anything toward embedding by the end of the summer.
....and that is a sort of no.
I can tell you one thing about the engine room: We will give every attention to support for the Jitter project, and we will immediately integrate and distribute anything that Ian produces in the area of higher performance.
More information about the Squeak-dev