[TCLCORE] The Remaining "Big Plan" TIPs...

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

Apart from the TIP vote currently running (if there are adjustments to
those TIPs to be done during the vote, send me what you think should
change) these are the things that I'm eyeing to do as part of the 8.6
feature set.

   TIP#162: IPv6 Sockets for Tcl
   TIP#180: Add a Megawidget Support Core Package
   TIP#234: Add Support For Zlib Compression
   TIP#244: PNG Photo Image Support for Tk
   TIP#281: Improvements in System Error Handling
   TIP#329: Try/Catch/Finally syntax

Now, for the constraints/notes. The big constraint is time: we want to
close the window on new features in early December so that we have time
to push the implementation into a fit state for an 8.6.0 release in the
(northern hemisphere) Spring. This means that some projects are going to
be necessarily much smaller than some people think.

For example, #162 (IPv6) is really not about the sort of major
re-engineering that David Gravereaux envisages, as that would take much
more work than we have time to do *even with an implementation ready for
one platform*. Given that, getting IPv6 support is actually the issue
before us and that's what I plan to call that TIP on. (Yes, a generic
plugin architecture for socket types would be neat, as would fixing the
blocking-resolver bug. But they're not this TIP.)

It's possible that #180 will be purely scripted, maybe not in Tk at all.
It certainly needs quite a bit of work to revise once I work out (with
Damon?) what is actually not already trivial. (TclOO changes the whole
equation quite a bit...) Do not be surprised if this slips and goes into
tklib instead.

TIP#234 has suffered badly from the bikeshed effect. For myself, I want
this sufficiently strongly that I'd be happy to eliminate chunks of the
Tcl API (we can do them in the future after all) and do just those parts
that we can be sure of. That then enables TIP#244, which is important
for Tk.

I'm inclined to characterize TIP#281 as more of an approval for what I
expect will be an ongoing programme to improve the error handling on
(especially) Windows. It's not clear to what extent we should make a new
API to support this yet; maybe in 8.6 it is sufficient to keep the added
functions internal for later exposure.

And finally, I'd like to see TIP#329 go in. Right now, doing stuff with
[catch], [switch], etc. is sufficiently messy and ugly that people
don't. This results in code that is not as robust. Yes, [try] adds
nothing that you can't do now. I expect the first version to be scripted
(though I hope to be able to produce a C version before 8.6.0). But the
key is that it much more strongly encourages people to do the right
thing with error handling (and hence error creation too). Doing the
other exception cases is purely a bonus from my perspective. (It does
benefit rather from #281...)

In order that we may run an orderly vote within our own deadline, the
"Big Ticket CFV" will probably be next Friday. If you want to argue
about the APIs in these TIPs, *please* get it over with before then! We
can't/won't wait for you to resolve some minor point!

Donal.