Re: Do we trust the man on the GC trigger?

[Dirk Herrmann]

On 28 Aug 2001, Michael Livshin wrote:

> one more thing: the `mtrigger' mechanism is, basically, a kludge.
> it's there simply because vector/string/etc. data is allocated by
> malloc (i.e. not on the Scheme heap), and thus some way is needed to
> account for that.

OK, I see your point:  The cleanest solution would be to have a more
general heap structure than we currently do, i. e. one that not only
stores cells, but arbitrarily sized memory regions as well, right?
This would allow for a cleaner separation of scheme and C memories.
There would probably be some communication issues if contents from one
domain were used in the other one (example: scm_take_string).

Best regards,
Dirk Herrmann