How often did you deal with mainframe BASIC code where 70% or more of it was remarks? Inside the loops? That paid by the K-LOC era wreaked havoc with performance in Basic, early DiBOL (before it offered compilation, what a happy day that was!)

And obviously I didn't mean languages that follow the p-code model. I would be hard pressed to call those interpreted, unless of course one also wants to consider Java and .NET to be interpreted languages.

Though there is fair debate to say that they are interpreted, not "virtual".

Did you ever have to optimize for the fact that loops -- Gee Whiz was horrible for this -- could see as much as a 40% speed increase just by making line 1 a gosub to your configuration and line 2 the start of the main loop?

Mentioning the 6809 is fun, I had a coco at home, and still have one in my collection. It’s cobbled together from different models for optimal “compatibility” and comfort in using. Rev NC board I did the 64k mod to manually

Coco 1 case, coco 2 keyboard… though I’m hoping to soon make a custom backplate so I can use kailh bronze switches… though deep down I’d kind of like to have a melted keyboard in it.

Got the “proper” original sticks for it now. When that pic was taken I was borrowing from my 1000SX. One might say I’m something of a collector.

I almost went "32k for an assembler? EDTASM fit on a 16k ROM. EDT/09 was 12k" but then you said with the source. As in the source of the assembler itself? That would be some tight coding indeed. Was it self building or did you need the "help" of an intermediate language?

Was it for coco/dragon, or a more esoteric beast like the SuperPet? Could it standalone to use all 64k, or was it reliant on system ROM?

Sorry, anything remotely related to the Coco gets me all happy.