As a last post in this blogathlon, what I’m going to do next: well, move, obviously. Also migrate my server stuff which is still running on old Dual P3’s to the new Supermicro server as soon as I get the new chassis delivered.

After that, next project should be opening a nlams2 point of presence for Phyxia Networks at the NIKHEF datacenter in Amsterdam. 

For the new home: finally get the entire mythtv installation in working order. This includes: DVB-T inputs for VRT and RTBF, DVB-S to Astra (we don’t have any sat hardware yet, that’s for after the move), dual analog cable tuner on the PVR500 card, all on the backend machine; then the correct setup of the living room frontend, building a new frontend for the bedroom (we’ll also need a monitor or TV for there) and also finishing my Java frontend so we can also watch livetv (and perhaps recordings) on our Windows workstations.

Enough work I’d say 😉 

Woop. You can’t own a number, plzkthx.

I’ve set up MythTV with a separate frontend and backend machine a while ago. Now, I don’t always watch TV on my TV, but also on my workstation (as does Diana), which runs Windows XP (yea yea, eww, eww, etc). There isn’t any working MythTV windows frontend available, and some people don’t seem to like the idea so they guess there never will be one.

I’ve asked a few questions on the #mythtv-users channel relating to documentation of the protocol, which is severely lacking, but apparently the only thing one should do in that case is bash the person who is asking the question. So I’ll just bash back: someone really needs to learn how to write a decent protocol.

For starters, there is no back/forwards compatibility. At the start of the session you announce your version, and if it doesn’t match with the backend, you just get thrown out. Done.

Then, it’s a bit of a ridiculous protocol, with 8 characters where you can put (in text format) the length of the upcoming message, and then for x characters, the message itself. Parameters are separated by []:[], okay, probably not a likely sequence to be somewhere inside a variable. But you can only do some things with this protocol. If you want to get information instead of action out of the backend, you have to connect to the MySQL database yourself, which means another open port, another grant statement into your MySQL users, and another way to get incompatible when some small part in the database structure changes.

MythWeather, the weather forecast plugin, is also broken. Well, it’s broken in reality right now because msnbc changed their site, but that’s bad luck and can be fixed. But it’s also broken by design, as there is a webpage part, in PHP, for your MythWeb server, and then there’s a python part for on your frontend. Both with totally different code, just pulling the location from the database – in my opinion the frontend should query the backend through the protocol for the current weather info and be done with it. A frontend should be just that, a dumb front end retrieving info from the backend.

MythMusic. Ah, another great piece of design. You have a music collection on the backend, which is set up for MythTV. Then you have your frontend, which contains the actual music player. However, the backend does not stream your music data to the frontend, instead you must mount a (smb/nfs/whatever) share on your frontend (on the same mountpoint as where the music is on your backend, no less) and then it will play your mp3’s locally.

Of course in the transcoder profiles, many ID’s are hardcoded so you can rename and switch all you want, the transcoder will still pick profile id 2 from your database wether it’s what you wanted or not.

I haven’t dug deeper into this stuff, but I’ll bet there’s more where that came from — in my view caused by lack of decent coder docs combined with a broken protocol implementation.

I read on the channel the backend would likely need a complete overhaul, the database structure is completely idiotic and the frontend theming sucks majorly, so one has to wonder if there is ANY piece of this software which actually works the way it’s supposed to. 

Update after more coding: it also crashes the entire backend when you give it bogus data. Combine this with absolutely no authentication whatsoever on the protocol and anyone can fuck up your scheduled recordings, tv watching, etc.

Update after calming down and receiving a wider audience by way of mythtvnews.com: many of these issues are known and are being worked on, but of course Rome wasn’t built in a day – if you can help, I guess it would be welcomed in making MythTV better — after this post you may think I hate it, but in fact I think it’s great, otherwise I wouldn’t be using it 🙂

While we’re learning Ruby and Rails at Syntra, Diana is also still studying Web Developer and is studying PHP. Apart from "obvious" small examples she recently made (with just a little help from me) her first "real" partial application, being a calendar like the one you see on the right here 🙂 In one of the books or tutorials, something similar was made but with lots more code and as far as I could see a good part more CPU load. Good job, schatteke! 😉

Good news! When writing software you can now certify your software, like the "Works with Windows Vista" label, with a "Works on my machine" certification!

At work, Mobistar delegated our reverse DNS to a zone with a slash in it. I didn’t even know this was legal in DNS but apparently it is, and BIND and PowerDNS serve it perfectly. I installed PowerAdmin to manage my PowerDNS records with a GUI (MySQL INSERT/UPDATE was becoming a bit tedious), but could not create a domain nor any entries below it (after creating the domain through MySQL), because it wouldn’t accept the / as a valid character. I cooked up a small patch (don’t know if it’s complete, I just know it works for me) which you can find here .

A funky dialog box from $program, courtesy of The Daily WTF:

 A bit harsh perhaps, no?

I was wondering why every time I did a search on ipkg.be the search counter incremented by 2. I checked through my code but nowhere was that statement being executed twice. After a while, thinking about something else related to AdSense, it dawned on me… I recalled seeing Google AdSense hits in my access_log for the site, requesting the exact same page as the user (to adjust to the content on the page). On a “search engine” ofcourse, this means that Google does the exact same search as the other user, effectively doubling the number of searches logged… I have adapted my code to check the user agent, will be fixing up the database later so that Google’s hits aren’t included anymore…

I realise it’s tuesday already, but I haven’t really found the time to write anything. So here’s a small writeup of Fosdem day two…

Woke up (too!) early, left at 9:30, arrived at 10:50. Still early enough to walk into RIFE, we thought. So we did. Very earl-y, as this guy decided to pronounce URL as “Earl” instead of “You are ell” 😉 The presentation wasn’t all that interesting, but I guess that’s what a Java Framework is supposed to be 🙁 Still, I would think it could have been presented more interestingly.

On to Subversion, a very interesting talk, which has opened my eyes even more about how SVN owns CVS in many ways. Also learned the O’Reilly SVN book is available for free (as in speech and beer) at www.svnbook.org.

Valgrind was next on the program, went maybe a bit too “deep” inside the matter, maybe assuming a bit more previous experience with valgrind or memcheck, but all in all not a bad presentation.

Lunchtime, we brought our own food as to not have to queue for 45min like last year 😉

Beagle presentation, by The Guy Who Now Works For Google. Very nice presentation, mostly questions and answers, but very interesting. Also, I was surprised at the quality of the questions asked. Useful and to the point. A total difference from what Rega people would ask, if they even asked anything.

OpenOffice.org, by Michael Meeks (who was preparing his slides right in front of me during the Valgrind presentation). Apparently mr. Valgrind was using a “magic build” of Michael for his presentation, which screwed with the colors somewhat *G*. What was very clear is that Michael was painting a picture of OOo being that it was not good at all, still way behind on what it should actually be, slow, etc. But the presentation was pretty humourous and light, showing new features in 2.0, showing what can and should be done, and ofcourse, a call for developers.

XUL. What can I say about that. The presentation absolutely sucked ass. The original presenter couldn’t come, so someone else came in his place. Assuming everyone knew what XUL was, and that even everyone in the room was developing XUL software day in day out, he enumerated directory structures, which files were in them, and then even the content of each and every file. Very strange presentation. When we left (after 20 minutes of almost falling asleep) , many others were too. We joined in OpenCA, and first thing I heard was “And this is the point of asses.” Must be Italian English for access 😉 10 minutes later, OpenCA presentation was over and noone seemed to have any questions, so we didn’t hear much of it.

Then, luckily we were already in the Janson room, it was Jeff Waugh’s turn. While waiting he put the Xgl demo video on the big screen for everyone to drool on. His talk was quite good. Design vs Megapixel! LAMP: Linux, Apache, Most of our scripting languages start with P, and PostgreSQL. More funny stuff. And very valid points made, iPod vs Zen, … A very interesting and nice talk to end Fosdem with.

Pity for the people manning the booths that everyone had packed up before that, as in my eyes this would be the perfect end of Fosdem, to go home and buy that one Debian T-shirt you first decided not to buy, or that O’Reilly book that you thought you didn’t want… not possible unfortunately.