On 6/20/07, Neal H. Walfield <neal@walfield.org> wrote:
At Wed, 20 Jun 2007 15:16:38 +0800,
Wei Shen wrote:
> I read the section. I am not quite clear about the storage of a file's
> translator attributes (I presume that a translator can not be associated
> with a file node via other ways than the *file_set_translater
> *interface),
Right.
> but I think it should be the responsibilty of filesystem servers to replace
> the translator path for a chroot process (in *file_set_translater* and *
> file_get_translator*).
How can it do this? It can't find out what the translated process has
access to.
Yes, you are right. I misunderstood previously and thought that the critique is worrying about the path of the passive translator is errorly interpreted but not the path argument to the translator.
I still think there are ways to solve this problem. For example, the fs server can add an virtual root argument to the passive translator, and the translator (which we trust) will later add this virtual root to any path argument provided by the chroot process.
Or, the fs server passes the virtual root of a chroot procees to the translator when it is actually invorked (by different clients). Of course, all such approaches need the awareness and cooperation of translators, which I am not sure if is too strict a requirement for them.
Regrads,
Wei