[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Systas and Guile
Re: Systas and Guile
Tue, 25 Sep 2001 14:20:35 -0700
On 25Sep2001 02:07PM (-0700), Tom Lord wrote:
> In recent messages, I've pointed out several features of Systas that
> ought to be in Guile. I've pointed out features planned in in both
> Systas and Guile, for which Systas provides a working starting point
> and a plan. I've pointed out ways in which the Systas code base is
> easier to maintain than Guile's. With the exception of the ()/#f
> issue, none of my suggestions have met with any objection.
I think you are reading too much into the lack of response. I would
guess the lack of response is for one or more of the following
* No one has the time to debate 20 totally disconnected issues at once;
it's much easier to consider issues of this sort one at a time. Thus
the greater response to the ()/#f issue.
* Your style of argumentation is highly dogmatic and often rude. This
makes discussion unpleasant and to some extent pointless since you
seem unlikely to change your mind about anything.
* Your descriptions of features are sometimes so vaguely stated that
they are impossible to discuss. For example, you have mentioned
"port/descriptor unification" and how it's realy great. But I don't
know what a descriptor is, in this context, or how it differs from a
port, because your definitions did not make the distinction clear. So
I can't intelligently comment on your suggestion. I suspect many
people are in the same boat.
* Some of the features you propose, such as mutable shared substrings,
were intentionally removed from Guile after the Guile/Systas split,
and the changes were discussed extensively at that time. People may
not want to rehash all of that discussion just because you weren't
here the first time around.
The one proposed change I recall there being some amount of support
for was switching to exact GC, but that is not implemented in Systas
and many people's support for it was predicated on a C preprocessor
that would hide the memory management details from C programmers.
> Were that all there was to it, the right thing to do would be obvious:
> adopt Systas as the next version of Guile.
It's up to the maintainers to decide that one, but I think this
suggestion is not well-founded. I hope others on the list will excuse
me if I am stating the obvious.
> But there is a complication: Guile has some features not found in
> Systas and it has interfaces depended upon by other projects.
It also has an ongoing stream of active development including
long-term projects like the VM that have not made it into the mainline
yet, and has been actively maintained for the past couple of
years. This development momentum is worth considerably more than any
specific set of features.