Slashdot Mirror

This is old news ;-) Back in May, baard.com reported on this.

I thought the Personal Server was so neat, I wrote my own article based on the baard.com info, Science Fiction becomes science-fact.

4. Processing of the Evil Bit

Devices such as firewalls MUST drop all inbound packets that have the
evil bit set. Packets with the evil bit off MUST NOT be dropped.
Dropped packets SHOULD be noted in the appropriate MIB variable.

# echo 0 > /proc/sys/net/ipv4/tcp_ecn

I can't take the credit for this discovery. I learned about it while sifting through the comments to this very article. Also, check this link for more info on ECN.

Unfortunatly, P2P is not really the same as radio. Lissining to radio means that you *want* a DJ to chose the songs for you. The truth is that there are fundamentally three parties involved in the music lissening process: the data provider, the DJ, and the lissener. These parties may be devided up in many diffrent ways. Examples:

1) "Data provider == DJ == Lissener" is the traditional mp3 and CD system.

2) "Data provider == DJ != Lissener" is the trditional radio system.

3) "Data provider != DJ == Lissener" is the fast download system you are describing.

4) "Data provider == Lissener != DJ" is the system used by my artificial intelegence based mp3 player Smartplay. It only playes songs from your drive, but it uses a simple AI to guess your mood and it had a more efficent user interface to help keep you from waisting a lot of time skipping songs.

Anyway, there is really nothing keeping us from a "Data provider != DJ != Lissener" system. This would mean that the DJs would broadcast URLs and mixing instructions, but not necissarily the mp3 data it's self, and the download sites would wait for a significant portion of the lisseners to login before starting the multicast download of the song. The point is that you would not need to have a lot of bandwidth to be a DJ and the lisseners computer can hack together the playlists from multiple DJs to allow for even greater variety.

I tend to think that a hybrid of the "Data provider != Lissener != DJ" and the artivifial intelegence selection of DJs (ala smartplay's selection of songs, but for whole DJs instead---based on what they were plaing right now and your mood) would be the very best solution.

Jeff

Actually, it's not just selection of media from the store which is a problem with CDs. It's selection of media for lissening. I wrote a little program called Smartplay which "tries" to do two things: (1) learn your moods to more effectivly chose songs to play for you and (2) minimize the ammount of time you spend chosing the music you want to lissen to (ala no long periods chosing playlists and no distractions from crappy songs comming up under random play).

It's buggy as hell, but it proves my point that a software mp3 player on a computer can be MUCH better then CD players and current hardware mp3 player.

The thing I do not understand is why would anyone *want* new search / format features to come from Napster. We should want cool AI based search programs and flexible music category systems, but we should want them to be independent of any specific music distribution channel. Napster and Mp3.com are about the 2nd to last groups of of people I'd want to help me figure out what I want (the RIAA, Sony, etc. being the last people). First, we have no reason to trust Napster's opinion. Second, a company will not really experement with diffrent methods like haphazard development of independent internet music locating services/programs will.

Now, once we forget about Napster and start tallking about AIs and music categories there is a lot of interesting stuff. I wrote a little Perl/GTK front end to mpg123 a few years ago which tries to learn your moods from your past lissening habits (SmartPlay). We could create a forum (web page) where independent musicians could claim to be simillar to famous artists, provide mp3s of their music, and lisseners would vote on the simillarity.

Anywho, there are lots of interesting things which we can do to make it easy to find new music, but Napster is not the company to do these things.

Product idea: a PCI card that renders MP3 to standard S/PDIF or optical that can be run directly into the back of your digital-enabled receiver.

Yes, that would be very cool. It's importent that the computer retain control of things like seeking with the MP3 and selecting songs. I do not want to be stuck to some fucking playlist or a massive random play. I wrote a little program which attempts to learn your music lissening moods. It's not very reliable or stable, but it makes the point that ALL current mp3 players and sterios have uninspired crap for user interfaces. Anyway, if Joe Random Geek has the ability to write his/her own music user interface then we should see some cool shit in a few years, but if the corporations keep giving us the 3 button remote control because they are affraid to cornfuse their customers then we will be stuck with the crappy interfaces we have now.

I must say I have been really disapointed with the tendency of OSS mp3 player to just copy winamp's skinability. The truth is skins are a really lame form of user configuration because they make it more difficult to change the interface in meaningful ways (as opposed to E themes or something where you actually do reconfigure the interface).

Now, you ask "who would want to change the interface to their mp3 player?" Well there are a lot of reasons, but they mostly boil down to (1) playlists are a waist of time and (2) you will waist you life away skipping songs you are not in the mood to hear if you use random play. I wrote an mpg123 front end called smartplay using perl/gtk which tries to fix these problems. Warning: this is ``proof of concept'' which means it will probable crash sometimes. Features: it uses random play but it displays the list of the next 20 songs it will play and allows you to remove them from the queue at any time, i.e. instead of "hitting skip on a crappy song, going back to work, getting another crappy song and needing to quit work (mentally) to skip it" you can just skip all the shit at once. Also, it tracks you lissening patterns to try and opimize it's random file selector to your mood as determined to the kinds of songs you have been lissening to compleatly.

Jeff

..is not what you lissen to, but how you lissen to it. If you have a large collection of MP3s then there is a good chance that you spend a LARGE amount of time skipping songs.

Example: Song you are not in the mood for comes on, so you stop work for a second and press skip and go back to work, but just before your brain switches back to code mode another song you don't want to here right now comes on and you must stop work again.

The tradiotnal solution to this is playlists, but it is easy to have too many MP3s to use them effectivly.. or just not understand your own lissening habits.

The solution I came up with is to use a primitive AI (well not really, but almost) to try and learn my lissening habits for me. It also shows you the next 20 songs it is going to play and allows you to cancel them from the list BEFORE they start playing.. this makes an incredible diffrence in the ammount of time you waist skipping songs in random play mode. You can check out the Perl source to smartplay, but be forewarned it is proof of concept.. and not really all that stable or polished. Plus, it takes a while to really learn anyhitng about you, but maybe someday someone who really knows something about AI will pick up the idea.


Related to efficency: There is music out there, like Brian Eno, which is specifically designed (well.. sorta) to make you more productive (well.. sorta). The idea being that the music removes destractions (well.. sorta). I personally lissen to Techno since it seems to fit in well with the mind set required for programming. If your a Techno hater you should try lissening to it while your programming.. I've seen people made into Techno fans this way.

Jeff

I wrote a little mp3 player wich uses something that is not a neural network (but could be replaced by one) to learn your music lissening patterns and adjust it's random number generator acordingly. It would be great if some CS type who knows something about real neural networks could write one into x11amp or something.

I was more concerned with the user interface issues of how to maximize the amount of information available to be learned then with the actual learning algorithm. Check out the code it's in Perl and uses Perl/GTK and mpg123.. and it will crash because it was only meant as proof of concept.

Jeff

Ok, I can understand that people would want some visual art with their music, but visual art could be distributed with MP3s too. Maybe some artists could even sell advertising by embeding HTML and related files in MP3 headers if the players would show the stuff. Regardless, a phyisical media is not the answer because it constrains the lissener too much. Why should I be forced to lissen to music in the order someone else prescribes or even in the random order prescribed by a large CD changer. Hell, I think playlists are even too constraining. I have writen a simple perl-GTK front end to mpg123 which uses a simple AI and attempts to learn the users moods (check it out here). It also allows you to cancel songs before you hear them which keeps the user from waisting lots of time lissening to the beginnings of songs they dont want to hear (unlike more random play options). Note: since the palyer is in perl it is easy to modify the AI with your own rules that consider things like artist or song name similarity. The point is that we to have the needed flexibility in players we need the player front end to be implemented in software and to be easy to modify. I never lissen to CDs not because they waist too much time.. my player is writen specifically to save me time.. that is the beauty of software. Jeff