The Smalltalk Revolution

Re:The Smalltalk Revolution by hga on Thursday January 08, 2015 @09:42PM (#8860445)

I would go so far as to say that the "MIT style" of using Lisp is file rather than image based. Saving and restoring an image is purely for minimizing startup time, and in development would only be done when you're at a point where that makes sense. And you'd start from a fresh band, LOAD all the relevant files, and then save. If you're doing an application, it might not make sense to ever go to the trouble.

Or so I remember as of the early '80s.

[ Parent | on HN ]

Re:The Smalltalk Revolution by lispm on Friday January 09, 2015 @11:18AM (#8862838)

On the Symbolics Lisp Machine it was common to create incremental Worlds. From an incremental image one could write a new incremental image.

For example I used to save a minimal site-configured incremental image with some OS patches first. From there I would load the software I use and dump the next incremental image. Into this image I would load more software and the stuff under development and make another incremental image. Based on this I would load additional patches as necessary...

The reason why these images were saved is not startup time. The reason was that loading code (-> applications) was quite slow on most Symbolics Lisp Machines during the 80s.

If you develop on a Symbolics Lisp Machine it first looks into the running image for many types of information, not the file system: documentation strings, arguments, who calls, ... even the documentation database needs its index to be loaded into the image.

[ Parent | on HN ]