Re: [Debconf-video] Testing Nageru in parallel during DC17? [was: (Re: Report from last week-end's single-user single-day DebConf Video sprint)]
On Tue, Jun 27, 2017 at 6:25 PM, Wouter Verhelst <w@uter.be> wrote:
> For clarity, I'm not just talking about performance; I'm also talking
> about the fact that mixing FullHD video on a single FullHD monitor is
> difficult, so you'd need to have two monitors, or a 4K one, or to scale
> your video as well as mix it. And then there's also the increased
> bandwidth requirements of FullHD as opposed to 720p.
When I saw Nageru in action, this was not an issue and preview frames
simply ran scaled in realtime on the same machine.
> It's definitely something we need to be considering; but given that 720p
> is already an improvement, not something urgent IMO.
Agreed; but jumping from SD to FHD, instead of HD ready, would even be
nicer given it's 2017 and even cheap laptops have FHD displays; so
people can consume this content.
> - We need to ship loads of hardware around the world all the time. For
> cameras this makes sense, since they are expensive to rent, but for
> laptops or computers (which can be borrowed or rented locally), not so
> much. If we want to use voctomix, we can just say "a reasonably recent
> processor". If we want to use Nageru, we need to say "a reasonably
> recent processor with a powerful enough display adapter that has a
> driver which is not crap". This is a slightly more problematic thing
> to do
I would argue that not shipping hardware around which needs N+2
redundancy and simply using off-the-shelf Blackmagic boxes you can buy
everywhere increases resilience. As to the laptops, anything
non-ultra-portable which is less than ~3 years old should suffice; and
those are easy to rent.
> - The team developing gstreamer and voctomix is larger than "just
> Sesse". We've just moved away from dvswitch that suffered from the
> "not enough developers" problem; we shouldn't move to another tool
> which might have or develop a similar problem down the road. Granted,
<re-arranged>
> I think nageru is awesome; when Sesse first blogged about it a few years
> ago, I also suggested it to the video team. But the response back then
> was, rightfully, that vocto seemed more mature. That bit is probably not
> the case anymore right now, but it's still true that we should be
> careful before we make the switch.
That is a valid argument, but code velocity still seems higher within
Nageru. As to maturity, I randomly remember
https://github.com/voc/voctomix/commit/b07239f8fe7a50939c52fff7fdb690861dfde916
from last year. So I would argue that vocto also has warts in places
where it shouldn't.
> - Nageru, given that it was written on top of Sesse's own video library,
> has less support for random hardware that we might want to use and/or
> connect to it. Granted, such support could be written, but that brings
> us back to the previous point.
I can't see Blackmagic going away; I get the point, but in the
foreseeable future, this seems not to actually be a constraint.
> I don't think 1080p50 (or, well, p60, given that we're dealing with
> Canada here, not the EU) is all that useful, but 1080p30 isn't something
> I'm opposed to -- in a few years or so, given the context of the above.
Given that these are talks, 1080p25/30 should suffice, yes.
> And no video this year -- that isn't an option. Remember, our current
> cameras are falling apart (and if they weren't, we're short a room).
>
> We need these cameras now, not in a few years.
Fully agreed.
I am not saying not to buy anything, I am merely saying that getting
cameras which fit the current profile of 720p might save cost now and
eg 4K cameras will be cheaper in a few years.
Though, as an aside, a mix of rental and buying seems to be more cost
effective than buying all of them, plus shipping and insurance,
anyway.
> We should definitely install nageru onto one (or more) of the machines,
> and test it out during off hours. But only during off hours; for this
> year, we should prioritize recording on voctomix, not nageru.
I agree with prioritizating vocto for DC17; the intention of this
email was to try and build consensus if it would be possible to have
Nageru run in parallel. If it's tested in the off hours, that's also a
lot better than having the video team in one place and not testing
this. Especially built-in monitoring should be a given no matter what
system is used, long-term. Being able to quickly and automagically
make at least basic statements about things not being on fire is
incredibly valuable.
The results would also be valuable for FOSDEM, I am sure.
>> I do know that this topic is controversial due to personal reasons,
>
> Not for me :-)
Aye; but it's still a possible point of contention and thus I wanted
to state it openly.
Richard
Reply to: