**> On Mon, 25 Jul 94 17:59:35 BST, dib@signal.dra.hmg.gb (David Bruce) said: David> A few people have tried to convince me to create a simple PC based David> implementation of Self and include it for free in a book for novices David> to spread out the word :-). I don't know if it is a good idea - Little David> Smalltalk and GNU Smalltalk leave many people with a bad impression David> of the language.

David> Just out of interest, what's the problem: functionality or efficiency? David> I'm not so familiar with {Little,GNU} Smalltalk; are they hopelessly David> `cut-down' versions of the real thing (cf. microemacs :-)), or just David> very very slow?

Both are, as I recall, hopelessly cut down. GNU Smalltalk includes an interface to the X Protocol but nothing more in terms of environment. Last I heard, Little Smalltalk was missing important things like instance variables :-)

David> I know that an efficient implementation of Self is decidedly non-trivial, David> but surely the point of the language is that it is simple and so could be David> implemented naively (e.g., as an interpreter) without an enormous effort David> (though not necessarily by a novice!).

An interpreter for the general ideas behind Self objects _is_ trivial -- a good hash table and a polymorphic "evaluate" function is most of it. Some pain, perhaps, to handle arguments. But most of the work would be in providing all the primitives and making sure that the standard libraries all worked (e.g., large integers, etc.).

David> This raises the following question (for instructors, I guess): David> how slow a system could one `get away with'?

Craig Chambers' thesis cited estimated speeds for a purely interpreted Self as beening about forty times slower than compiled C code. Is this acceptable? Eh, I don't know. Certainly, we've never had a system fast enough to make us really happy for instruction....

Brook