Re: [TCLCORE] Pre-CFV: TIP#257

[Donal K. F. <don...@ma...>]

Gustaf Neumann wrote:
> Bring the full flexibility to the oo level
> is a logical step.

I notice that it doesn't propose to allow control over the dispatch
engine. (If you're going for full flexibility, you might as well go for
*full* flexibility! :-))

> Tip #279 goes even one step further and provides an multi-flavor
> oo framework.

I think it's lightweight on detail. For this audience, detail is very
important since we know that's where the devil resides.

> I still believe that the framework aspect is important
> since it allows continuous work in extensions for flavours,
> while having a stable part in the core.

Of the mechanisms actually described in the TIP[*], the setrelation
command has a problem in that the relations themselves are a limited set
(which comes back to the point about dispatch engines). Moreover, the
alias command makes some dodgy assumptions about the implementations of
a method (e.g. that the instance variables are meaningfully scopeable to
the command, which really is dodgy given how Tcl commands work). There's
also no proposed C API, which would cause problems with adoption by some
parts of the Tcl community (not myself BTW).

BTW, while I'm thinking about C APIs and independently of anything
else... is there any documentation of the XOTcl C API? I couldn't find
any in the 1.6.0 download bundle. Has it been omitted from the bundle
build process by mistake? (Also, the bundle I downloaded from SF seems
to be missing the HTML version of the language reference, and that
really does look like a bundling bug.)

Donal.
[* This comment applies specifically to Revision 1.9 of the document. ]