[EROS-Arch] Questioning need for Call Count
Sun, 12 Nov 2000 20:06:48 -0800
At 12:24 AM -0800 11/12/00, Seth Arnold wrote:
>* Jonathan S. Shapiro <firstname.lastname@example.org> [001109 10:54]:
>> [...] An alternative would be to build room for a call count into the
>> data exchange protocol but not implement call counts within the kernel
>> mechanism. [...]
>Would the space for this call counter come from one the four
>capabilities registers passed in IPC communications? (Furthermore, is
>there any magic reason for passing four capabilities? :)
As the originator of "four" in this context let me answer that we saw magic in
hardware structures with fixed numbers of fields, such as definite
word length, definite number of registers. It was the style of efficient
machinery and we viewed the kernel as an extension of the machine.
I understand that many find this magic insufficient.
The 4KB string length limit was a place to stop short of promissing to
pass more data than would fit in core atomically. Short of an entire
address space it seemed that protocols would already be required
for passing more than the kernel's limit and the question was then
not whether to have a limit but the engineering issue of how big to
make the limit. Charlie Landau has had recent thoughts on
non-atomic passing of an unlimited stream.
Norman Hardy <http://cap-lore.com/>