Sunday, January 09, 2011

XBMC vs Boxee vs file browsers

I ditched Boxee the other day. I was sick of it eating far too many resources, never quite keeping up with my mouse, fighting me with strange user interface metaphors; ignoring some files it would never recognize and providing me with not much enjoyment.

I installed XMBC and got rid of boxee. That was before I realized how good I had it: Boxee's scraper doesn't work off regexp and demand you do crazy things to your filesystem.

What I am struggling to understand is why I can't simply have a damned metadata layer and scraper for my desktop which plugs easily into the file browser.

All that both Boxee and XBMC are really doing is scrape, fetch, and organise. XBMC does it better with a few widgets like "recently discovered espisodes"; but even then; how hard is it?

I advocate the semantic web stack just because this is the very problem those tools should be able to solve in a trivial fashion; but I don't understand why this has not been achieved by anyone else in the last decade.

It's not like web scraping is mind bogglingly hard, nor matching little bits of string against certain sources. Musicbrainz solved the problem for mp3s some time ago: if the same can't be done for video content, I would be surprised. Additionally, it's not like relational databases are a new thing: they have been around for a bit, and that's all you really need: schema, inserts, and the web scraper organising it all.

So it really, really pains me when the Boxee/XBMC/Gnome world can't get it right. Why does Gnome/Gnautilus still treat media basically like this:

Whilst your original design might have been GUI filesystem explorer, what I actually want from you is GUI-like-XBMC-Metadata-displayer.


den parser said...

That's great.. but still I choose file browsers for simple way.

Escalante Bloggers

Jay said...

Hi. I've used XBMC Live for the last 2 year, and the scraping features seem to be the weakest limitation of XBMC. They work for a while, but when say upgrades their website, the scrapers break. This is inevitable with scrapers. But trying to get new versions of the scrapers is a nightmare! If you're game to rewrite the xml scraper files yourself, you can, but you have to learn regular expressions, xml, character encoding etc, then you have to work out how to test them, apply them to your HTTP, all without much useful input from the XBMC developer community. Web scrapers are an essential part of a HTPC. Without scrapers (or some other movie/tv show metadata lookup system) you don't have a movie library (with plots, genres, years, actors, directors etc.) - you only have a file list and a media player. Whoopee doo!

You'd think an established community with as many users as XBMC could get their act together and post a link where to download the latest scrapers. Noooo chance. Try searching for the latest IMDB.xml scraper for XBMC 9.11 (for example) on Google or on the website/forums. I'll give anyone $100 if you can find a link to the latest working scraper update.

It's such core functionality, it's truly bizarre that it's so hard to find scraper updates. I'm considering moving away from XBMC now. I'm too busy to spend weeks rewriting scrapers myself. There should be a community to collaborate on updating scrapers.

Without a fully cross-referenced library, my system is just a file browser/player. I'm considering Boxee now. How does Boxee handle scraping media metadata? When websites like change, how long does it take for Boxee to upgrade their scrapers? How hard is it to find new scrapers?

I'd appreciate anyone's thoughts on this matter. I'm so disappointed with the attitude of some of the XBMC developers (I'm a software engineer myself, so believe me, I'm naturally sympathetic) that I would be happy paying good money for HTPC software that works.


Dan said...

@Jason, since I wrote this originally I've discovered Tracker - a gnome metadata layer.

I'd love to see scrapers which plugged into that instead of xbmc or boxee - so my filesystem knew what a digital object was.

Pumping that info back into xbmc would then be fairly trivial. I did some work with combining the xbmc database and tracker database; but sadly there are few libraries which can encode metadata in common video formats available.