nasko
2017/04/21 00:17:32
I know dcheng@ also commented on this, but what if
I know dcheng@ also commented on this, but what if we start out with an ID
generated in the browser process? If it needs to be calculated (based on origin
and such), then we should evaluate whether we have all the information in the
browser process. If we don't, then we are missing information and we likely
should have it, but if at that point we need the renderer process generating it,
it can be done at the time needed with the exact requirements and limitations
clearly articulated.
oystein (OOO til 10th of July)
2017/04/21 17:57:55
I'm slightly confused by this; what's the suggesti
On 2017/04/21 at 00:17:32, nasko wrote:
> I know dcheng@ also commented on this, but what if we start out with an ID
generated in the browser process? If it needs to be calculated (based on origin
and such), then we should evaluate whether we have all the information in the
browser process. If we don't, then we are missing information and we likely
should have it, but if at that point we need the renderer process generating it,
it can be done at the time needed with the exact requirements and limitations
clearly articulated.
I'm slightly confused by this; what's the suggestion exactly? In the current
prototype CL I've got the manifests set up in such a way that only the browser
process can call CreateCoordinationUnit; if a renderer process wants a
messagepipe to a CU, it has to bounce through the browser process. So the in the
cases where we don't want auto-generated IDs, the browser process can either
construct it itself, or validate it at that point.
nasko
2017/04/24 22:10:50
This is where my lack of full understanding of man
On 2017/04/21 17:57:55, oystein wrote:
> On 2017/04/21 at 00:17:32, nasko wrote:
> > I know dcheng@ also commented on this, but what if we start out with an ID
> generated in the browser process? If it needs to be calculated (based on
origin
> and such), then we should evaluate whether we have all the information in the
> browser process. If we don't, then we are missing information and we likely
> should have it, but if at that point we need the renderer process generating
it,
> it can be done at the time needed with the exact requirements and limitations
> clearly articulated.
>
> I'm slightly confused by this; what's the suggestion exactly? In the current
> prototype CL I've got the manifests set up in such a way that only the browser
> process can call CreateCoordinationUnit;
This is where my lack of full understanding of manifests is showing. How is this
actually enforced? Looking at the manifest file, it says coordination_unit is
provided by resource_coordinator::mojom::CoordinationUnitProvider. However, it
doesn't specify any restrictions on which process can call this.
> if a renderer process wants a
> messagepipe to a CU, it has to bounce through the browser process. So the in
the
> cases where we don't want auto-generated IDs, the browser process can either
> construct it itself, or validate it at that point.
Ken Rockot(use gerrit already)
2017/04/24 22:17:56
Services do not specify who is allowed to request
On 2017/04/24 at 22:10:50, nasko wrote:
> On 2017/04/21 17:57:55, oystein wrote:
> > On 2017/04/21 at 00:17:32, nasko wrote:
> > > I know dcheng@ also commented on this, but what if we start out with an ID
> > generated in the browser process? If it needs to be calculated (based on
origin
> > and such), then we should evaluate whether we have all the information in
the
> > browser process. If we don't, then we are missing information and we likely
> > should have it, but if at that point we need the renderer process generating
it,
> > it can be done at the time needed with the exact requirements and
limitations
> > clearly articulated.
> >
> > I'm slightly confused by this; what's the suggestion exactly? In the current
> > prototype CL I've got the manifests set up in such a way that only the
browser
> > process can call CreateCoordinationUnit;
>
> This is where my lack of full understanding of manifests is showing. How is
this actually enforced? Looking at the manifest file, it says coordination_unit
is provided by resource_coordinator::mojom::CoordinationUnitProvider. However,
it doesn't specify any restrictions on which process can call this.
Services do not specify who is allowed to request their capabilities. That would
necessitate layering violations. Instead they simply name their capabilities so
that any other dependent service in the system may "require" them by name.
The Service Manager enforces the static intersection of provided and required
capabilities. It's up to security review to decide whether one service requiring
any given capability from another should be allowed.
>
> > if a renderer process wants a
> > messagepipe to a CU, it has to bounce through the browser process. So the in
the
> > cases where we don't want auto-generated IDs, the browser process can either
> > construct it itself, or validate it at that point.
Issue 2798713002: Global Resource Coordinator: Basic service internals
(Closed)
Created 3 years, 8 months ago by oystein (OOO til 10th of July)
Modified 3 years, 7 months ago
Reviewers: chrisha, dcheng, fmeawad, jam, nasko, Primiano Tucci (use gerrit), Ken Rockot(use gerrit already), Sami
Base URL:
Comments: 130