On Wed, 2013-07-17 at 12:26 +0200, David Henningsson wrote:
> On 07/17/2013 11:22 AM, Tanu Kaskinen wrote:
> > On Wed, 2013-07-17 at 09:27 +0200, David Henningsson wrote:
> >> On 07/16/2013 03:20 PM, Tanu Kaskinen wrote:
> >>> Yes, there can be both an explicit and a default connection between two
> >>> nodes, or it can be also thought as being one connection that is both
> >>> explicit and default. I don't know what is a better way to think about
> >>> it, but perhaps it doesn't matter anyway.
> >>
> >> I would prefer the latter way. I would then see "default" and "explicit"
> >> as a property of the edge.
> >
> > I agree about that, at a conceptual level. Would you perhaps like to
> > have it this way also at the code level, i.e. have connection objects
> > with "default" and "explicit" fields instead of storing the information
> > in the nodes?
>
> Good question. Actually one option could be to store the connections as
> global list-like matrix instead of lists on the individual node objects.
> After all, you'll need the connection on both sides, and unless you want
> to duplicate information, you'll have to do a global search on one of
> the sides anyway. Since the number of nodes < 100, I don't think this
> would matter from a performance standpoint.
>
> You would end up with a global list with entries looking something like:
>
> struct pa_node_connection_t {
> pa_node *from_node;
> pa_node *to_node;
> bool default;
A single "default" boolean isn't enough. A connection may be part of
both endpoints' default routing, or just one. It's necessary to
differentiate between the different cases, because disabling one node's
default connections may not remove the connection, if it's also a
default connection of the other node.
> bool explicit;
> bool explicitly_disabled;
> };
>
> Where there is only one entry for every from_node/to_node combination,
> explicit and explicitly_disabled cannot both be true, and there is a
> connection if and only if (!explictitly_disabled) && (explicit || default).
> And no list entry for a node combination, is the same as all three bools
> being false.
I don't like the idea that there would be a connection object also for
connections that don't exist. If the "explicitly disabled" feature is
needed (I will argue later in this mail why it's not needed), it can be
implemented in some other way, such as a separate blacklist of
connections.
I do, however, think that having separate connection objects is a good
idea after all. I think it's entirely possible that we would at some
point want to have additional properties for connections (I don't have
any examples in mind, though), or we might want to refer to a connection
with some more convenient way than with a pair of nodes.
> >>> I don't like the idea that if a client wants to move a stream, it has to
> >>> break the operation down to "remove + add", and then the server tries to
> >>> guess what the client really meant. Having a "move" operations keeps the
> >>> client intention obvious.
> >>
> >> Fair point, although the move can be on both sides, whereas one type of
> >> move would be "I want node A to take data from C instead of B" and the
> >> other type would be "I want node A to output data to node C instead of B".
> >
> > These both are served by the move operation.
>
> In the RFC you only have pa_context_move_node_connection, which serves
> the latter case only. The former case is from "C -> A" to "B -> A", the
> typical example would be a client wanting to move a recording stream
> from one source to another.
If you think so, then it's a bug in the documentation. I really meant to
support both cases with pa_context_move_node_connection().
> >> Or perhaps you want to swap, so that if you're currently on "A -> B and
> >> C -> D" and you want "A -> D and B -> C" and you want to do all of this
> >> atomically. The number of operations you want to do might grow out of
> >> hand unless you have a "batch mode".
> >
> > The core doesn't currently have a concept of atomic swapping, so I don't
> > see the need for a separate swap command, but the core has a concept of
> > atomic moving.
>
> Well, IIRC, we have move_start and move_finish commands? A swap command
> would just be two move_starts and then two move_finish.
Good point. You got me convinced :) So, let's forget about the separate
move operation, and perhaps also the separate add and remove operations?
pa_context_edit_connections() should be good enough for everything.
> >>> What operations do you mean? Moving or removing a default connection is
> >>> not supported as such, but if the client tries that anyway, we can
> >>> implicitly convert the connection to an explicit one and disable default
> >>> connections, or we can require the client to do these operations
> >>> explicitly, but I think the latter would be too inconvenient for the
> >>> client.
> >>
> >> Ok, I think I didn't read the proposal well enough. Having done that, I
> >> understand that you're suggesting a global switch "default connections
> >> on/off" only. Or is it a per-node switch?
> >
> > It is a per-node switch.
> >
> >> I have another idea that might be worth considering: how about that the
> >> "explicit" layer can both enable and disable connections? So that there
> >> could be a default connection between A and B, but there is also some
> >> sort of explicit override that disables it. This would be more flexible
> >> than a more global on/off switch.
> >
> > I'm not sure what you mean. Do you perhaps mean that the default
> > connection on/off switch should be per-node (which it already is in my
> > proposal), or that it should be per-connection (so that if there are
> > multiple default from node A, it's possible to disable only a subset of
> > those)?
> >
> > I didn't make make it possible to disable individual default
> > connections, because I had a feeling that it would have very messy
> > semantics. If default connection from A to B is disabled, what is the
> > routing code supposed to do when conditions change and the default
> > routing is re-evaluated? Can it ever reactivate the connection between A
> > and B again? Is the per-connection disabling handled as a blacklist of
> > connections that must never be automatically activated?
>
> If the A -> B route is explicitly_disabled, that overrides any default
> connections the routing system tries to make.
What is the use case for explicitly disabled connections? I'll assume
here that your idea was to allow moving a default connection elsewhere
(making the connection explicit in the process) without disabling all
default connections for the node.
When the user moves a default connection, the routing system obviously
shouldn't immediately create another default connection elsewhere to
replace the disabled connection.
On the other hand, if the routing system doesn't create replacement
connections, then that results in weird behaviour. Let's say that
there's a default connection A -> B, and the user moves the connection
to A -> C. Then B disappears. The routing changes its opinion of the
best available routing for A, which might be D. So removing node B
resulted in audio suddenly appearing in node D, even though nothing was
playing to B.
--
Tanu
