On 24-05-2013 18:28, Casey Ransberger wrote:
> Sorry about the slow reply. I assumed the best intentions (hence the
> smiley face) but wanted to make sure folks knew there was more than
> one contributor (we have a whole list.)
You're right. And as I told, I like the Cuis idea. And I'm happy to see
that there's a community caring of it. From my side I'll look much
closer at it (I have used it a little bit, but current projects both
professional & academic swallowed time to experiment around).
>
> As for your "Fermi barrier of complexity" well. I don't know exactly
> how you intended this metaphor.
Things tend to be stable at a certain energy level. In order to send
something from a stable energy level to the next one it is usually not
enough to increase energy continuously. It is required to have a large
energy input and then stuff jumps to next stable level. Otherwise it
will find an unstable level and eventually decay to original place
emitting radiation. In academy radiation is good (papers, book chapters,
etc) but in business it is usually bad.
>
> (...)
>
> More importantly, I think, is that Cuis folks want to be a part of the
> Squeak community. We're a *distribution*.
I enjoy the idea of distribution. Clean & small & consistent distribution.
>
> Anyhoo, I figured I'd look at some numbers here. Someone should
> correct my incantations here, because these numbers are just too good
> to believe! I must be doing something wrong.
>
> So I'm tallying classes and traits in Squeak and Pharo with:
>
> Smalltalk allClassesAndTraits size
>
> ...and in Cuis, since we don't have Traits, with:
>
> Smalltalk allClasses size
>
> I realize that Traits aren't exactly classes, and that this is overall
> a relatively weak single metric for complexity, but just to show the
> difference in magnitude, please bear with me. FWIW, Pharo, which uses
> Traits extensively, only has  85 of them AFAICT, so they shouldn't
> skew the numbers too much.
>
> Squeak 4.4:   2148
> Pharo 2.0:      3301
> Cuis 4.1:         655
>
> I think just from that it's easy to see the signal:noise ratio in
> Cuis, but of course there's plenty of room for a deeper study and more
> metrics (cyclomatic complexity, etcetera.) One useful insight might
> be: Squeak's core should be able to get as small or nearly as small as
> Cuis is now (at *least.*) I think this is good news all around.
>
>
I agree with aiming to make Squeak core about same size of Cuis.