I'm trying to keep a todo list around of what we need to do at work.
I've started thinking about the user interface, and have translated one of the use cases to a scenario. It's enlightening to force yourself to actually try coming up with a UI, because it adds a bunch of design restrictions and forces you to cover your ground again.
I'm happy we're using Python and that we have Johan around. He's fixing bugs in PyGTK as we encounter them and I can bother him nonstop with all sorts of newbie PyGTK questions. Making GObjects in Python is insanely easy compared to the equivalent in C. So much so that you just create a GObject when you need it, because you can rely on its signal handling working well, whereas in C you'd hack around it.
So I've written a UI for a producer using a test signal, and a UI for a preview window for video that you'd be able to insert in any of the components on the server to see what's happening. Now back to the drawing board to incorporate things I need that I've figured out with these tests.
Need to stresstest the server regularly as well. I've found a bunch of HTTP stress tests but they focus mainly on how well a server processes requests. We're looking for something that would either allow us to "capture" requests together with their low-level read/write times and then play them back. The difference with HTTP request replayers is that we really want to replay the specific read/write behaviour, since this also effectively emulates both the speed the client reads at, as well as the specific burstiness of the client.
Or, any project that can generate requests this way, where you are in control over the simulated clients' speed and their specific request profile would be good. Drop me a note if you know of a good one.
I'm trying to keep a todo list around of what we need to do at work. I've started thinking about the user interface, and have translated one of the use...
Doctor Edgar Villanueva, for your viewing pleasure, with live translation. I have to say I was sort of disappointed about Bruce Perens acting like a suit and upstaging him and basically reading out loud a huge question. It wasn't necessarily wrong or bad to bring up the point that using laws to force Free Software might not be the proper solution. But making it into a little show felt like the wrong way to do it. By contrast, the Doctor gave a very lucid answer, well articulated, clearly showing that this guy has very good ideas and expresses them very well too.
I hope he learns English, so he can make his talks even more interesting.
And I hope next year we do live translation subtitling :)
Doctor Edgar Villanueva, for your viewing pleasure, with live translation. I have to say I was sort of disappointed about Bruce Perens acting like a suit and upstaging him and...
Day two of GUADEC, setting up stuff in the morning is really smooth now. Though getting up at 7.30 still sucks. Next year we have to have some other people actually do the streaming using our stuff, so we can hang out and sleep :) We even had time to make sure we had the sound correctly, and the wicked parabolic mike to get audience questions should work now too.
Lots of supportive comments from everyone both on and off the GUADEC grounds. One guy mailed me with a screenshot saying "Yes, the stream is working and simply rocks! I felt like beeing there."
That's exactly what we're aiming for. This is our first time streaming video, and we've already taken notes for everything we can do better next time (especially, getting high-res slides in the stream, for one thing), but it's very motivating to spend time discussing "what sort of rocking things will we have ready for the next time" instead of "oops, what did we do wrong ?". We made it. All the while our little server that could was running close to 3% CPU. Hm, we need a bigger testcase I guess.
Some random statistics: yesterday's viewer's ratings high score were achieved by none other than Dave Camp, who had 60 people watching, and people asking questions for him on IRC which were proxied through some of the gnome hackers.
Given a stream roughly close to 60 kbyte/sec, that accounts for almost 29 mbit/sec. The server is on a 100 mbit/sec. I hope nobody here on the university minded much.
Day two of GUADEC, setting up stuff in the morning is really smooth now. Though getting up at 7.30 still sucks. Next year we have to have some other people...
The server project is just a *lot* of fun to hack on.
Here's
Wim yesterday fixing the overlay code and grabbing the logo.
Yesterday night's new code, today in production. Yes, we don't know what we're doing.
One more thing left to correct that's probably again due to caps negotiation (anything with more than three elements just needs fixed caps inbetween all over, this has to be fixed), and then I'm not touching the machines anymore.
The server project is just a *lot* of fun to hack on. Here's Wim yesterday fixing the overlay code and grabbing the logo. Yesterday night's new code, today in production....