zvrba at globalnet.hr
Tue Sep 6 16:52:55 CEST 2005
'Lionel Elie Mamane' wrote:
> Please do so. I'm curious how you will handle:
> 1) Pointers being passed
> By copying the whole address space back and forth at each call and
> return? "Morally" that's not running in separate address spaces!
Make the programs share their _data_ segments, but NOT their _code_
segments. GPL is about _code_, not about the _data_ created and used by
The first pointer-sharing reference will trigger a fault which will be
handled by the ld.so and it will create an appropriate data-sharing mapping.
"Morally" I don't care, even in the case of copying.
Technically, making the programs share their data segments is, IMHO, not
a violation of GPL since no code is shared between the programs. But
then again, I'm not a lawyer. I'm always interested in other opinions.
> 2) A library that calls exec() or fork() or setuid() such a
> "process state changing" syscall.
> I don't think you can keep the semantics of all libraries in this
By providing the support for few such critical things within ld.so, so
it can change the state of the 'right' process.
I don't have all details worked-out, but none of them seem really
unsurmountable. In the extreme case, nothing that couldn't be solved
with little kernel-side work and support.
> It would certainly be a fun legal challenge. I don't believe however,
> you would win it. But I'm not a lawyer, he.
Neither am I.
> By all means, please follow through on this plan. It will be very fun
> to watch!
In what way "fun"? :) I don't have the time currently, sadly. But, the
biggest journey begins with the first step, so I just might start to
write some design paper in my spare time :)
Either I'll be the first one to do it, or someone will get ahead of me.
In any case, Werner will run out of his only reasonable argument (IMHO)
for not supporting PKCS#11 and users will (hopefully) profit ;)
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 254 bytes
Desc: OpenPGP digital signature
Url : /pipermail/attachments/20050906/a38657a3/signature.pgp
More information about the Gnupg-devel