Friday, September 30, 2005

Weekend Involgo Plans

With some luck and determination, I'll be able to make some good progress over the weekend. I've had overtime all week, and I still have to come in for a full shift on Saturday, but I'll have some time before work then and again on Sunday, and I'll have a prototype of my Involgo database working by the end of the weekend, I believe. Depending on the exact conditions of the prototype at that point, I may or may not be posting a link to an initial alpha release. We'll see.

Wednesday, September 28, 2005

Will XForms eliminate XUL?

Will the introduction of a complete XForms implementation in the Mozilla platform eliminate the need for XUL? I wouldn't be surprised if the current implementation uses XUL for the actual rendering of the forms, but wouldn't it make sense to work toward eliminating XUL from the Mozilla platform, in favor of a standard alternative? On a related note, it seems there could be plans to incorporate OpenGL or some similar API to the new canvas element, which would mean that if you allowed document rendering to a canvas element (XML is finally looking cool) you could have the entire window just be a big canvas element, render the browser's own XForm document to it (which would render the pages in their own canvas elements!) and all of this, being drawn with a clean and proper API, could be hardware accelerated, where such is available on Mac OS X, Windows with Avalon, and decent configurations of X.org.

Sunday, September 25, 2005

Pandora Take Two

Since my last post about the service, I got a comment from Tom Conrad, the CTO of Pandora. I've had more time to think about the service, its good qualities, and its drawbacks. I'm sure Tom keeps a custom blog aggregator to find anyone posting about Pandora, but I'm still honored for that to be my first post (not counting the Blog Comment Spam I had to delete). I'm surprised I don't see any posts by Tom on Ned's blog, though, I guess Tom just peeks around on Blogspot and doesn't catch the independents.

Anyway, Tom did say to feel free to continue blogging about Pandora, which I will do, because I have some more thoughts I'd like to share, especially knowing that he is listening.

Pandora is great. Its the best online music service I've found, and possible the greatest source of music I have, next to listening to Nights with Alice Cooper on the nights at work when I get the desk with the radio. When I say something is the best I've found, that doesn't really say a lot. It does in Pandora's case, but I'll clear it all up here.

What Pandora does right:
  • Deployment as a web app
  • Multiple stations
  • More fine grained than genre and year
  • Finds new music
  • Dirt cheap subscription ($36/year or $12/quarter)
What Pandora does wrong:
  • Can't listen by genre and/or year
  • Think computers are the best at organizing music
  • Pass up on the greatest source of music listening habits since iTunes. Maybe better than iTunes.
The last point is where I'm going, so bare with me. Let me go over the good in more detail. I'm giddy over the pricing. I can't believe people really pay $.99 a song from Apple, and I'm sickened that the music execs actually think that isn't enough.

I love when Pandora plays a song, and I think, "Hey, this is pretty good, who are these guys?" Of course, I also get Tivo flashbacks when I think, "Hey, why are you playing that?" It does find some nice tunes and I would be out of my league to acquire and organize a playlist of that quality and size. Now, I'm a Linux user (though more and more distastefully), so I'm happy Pandora went the web app route. It makes it so much easier to refer the service to my friends, who can just go to www.pandora.com and start listening, just like that. And, thank god they don't just group by year and genre, because I might be in the mood often for 80's music, but lots of music not made in the 80's will fit that class of music.

Sometimes, I just want to listen to 80's music, and how do I tell Pandora that? I can't seem to search for shared stations by name so I can just find anyone's "80's Music Mix" station. The redefinition of the very meaning of a station is a god-send, I must say. I can listen to the same station as thousands of people, and skip over any song I don't like. That freaking rules, because I hate when the radio plays something I don't like.

Now, what is the biggest thing that Pandora is missing? First of all, let me say this is speculation on some level, because if they take advantage of this, it must be completely on the internal level. They have thousands of listeners voting songs in and out of various groupings and all of this is stored into a wonderful database of golden information. So, Tom Conrad, I ask you, where is the "Play what other's like to listen to with this" option? The Music Genome Project may be a great venture and successful at what it does, but nothing beats a fellow human with a common taste in music. How about mining that database for common listening patterns, generating collective stations?

I'm sure they may find this a scary idea, because in essence it seems to go against the very beginnings and roots of Pandora itself, but it is also quite possibly the biggest value in the service for the company behind it, and to not take advantage of this would be, to put it simply, kind of stupid. (Sorry, Tom)

C#, .Net, and Mono

So, I've been doing some tutorials on MSDN on the C# language. It seems like Java without the emphasis on a virtual machine. Or, like C++ with emphasis on portability of useful code. Either way, the gun is in a lock box instead of just having the safety on or being taken away completely. I'm going to be looking into Python.Net and IronPython very soon, as well, and the Boo programming language, which is a Python-ish .Net language.

So, from the get-go, I admit that I do like C# for what it is. I've been looking for a compiled language for those cases when I would care to have one, and maybe it would be a good way to diversify my skill set. I've even considered an eventual implementation of Involgo in C#. With Python.Net I can wrap a .Net assembly for Involgo into the original Python API, and let any existing code still work, but likely faster.

C# looks like it would let me do a lot of the things that I like Python for, but without making them too difficult to manage like some other compiled languages. Delegates are nifty for passing function-like objects. I can pass objects without a specific type when I really need to, or have static typing when I need/want/think I want it. And, the automatic usability from so many languages is just "to die for".

I do wonder, however, what the consensus among the .Net community is for the Mono project. Most .Net users are definitely Microsoft users, so what do they think about us "others"? In the FS/OS world, the community is king. In their world, the community is surely strong, but it is far different. I wonder often about this rift between geeks. We're all after the same thing, we just think there are different ways to go about it.

Saturday, September 24, 2005

Lepers in America

Before Leprosy was cured, all the lepers would be sent to live together, away from the rest of the world. It might seem cruel by todays standards, but a horrible, killing disease is what it is. And, someone infected might as well be that disease walking.

Today we have several horrible diseases plaguing the planet. We have HIV spreading across the globe, for example. We don't put people with HIV into special camps. No, instead we just tell them, "remember to wrap it before you tap it." Meanwhile, the little AIDS demons continue to laugh at the world they may some day cripple.

Now, I'm not suggesting that we lock every with an STD up, but shouldn't there be some accountability? If you are walking around able to cause such an ammount of death and destuction, you'd think someone would say, "Hey, you better own up to this."

Crazy people are still punished for killing people, but is it a crime to know you have HIV and sleep around anyway? If you knock someone up, you have to support the child, but do you have to pay for their HIV medications if you infect them? People in this country won't even where protective masks when they have a bad cold, so why should I expect anything reasonable?

Pandora makes me want to spend money on music again

I love music, and I hate to buy a CD. To think how much I'm paying for that CD, with the full knowledge of how much it really costs to make, and how much the bands I love even get from the sales. I'd rather buy Ogg Vorbis streams directly from the band members, but that won't happen while the RIAA is still running this country.

Still, Pandora has given me the desire to spend money on music again, by giving me actual value in its service. Thanks, Ned, for blogging about this.

Is there room in the bubble?

There never really was 'a' bubble, and so 'the' bubble didn't really burst. There were bubbles, and there were a lot of them and a lot of them got big, and a lot of them burst. There are still a lot of bubbles.

Do any of these bubbles have room for me? Can I make my own little bubble and hope I don't pop it myself?

I'm evaluating my options for a start-up. I'm considering costs, estimating revenues, and trying to decide the timetable in which I could make it actually work. I'm doing my best to balence a quick turn around with a solid investment of my money and time.

Lots of geeks have done this, and I'd like to think I could be one of them who succeeds. And, I do think that I will, too.

Thursday, September 22, 2005

Involgo Progress from Scratch

I made plans, I designed things, and I spent lots of time making very little progress. Finally, thanks to a suggestion on IRC, I just dove in, scrapping the ill-fating code I already had, and just making things work. My little object database now works, kind of. It doesn't store data between sessions, it just keeps them as Python objects in memory. I'm able to easily and quickly add in all the features, debug the database, and iron out my ideas. When the major features are generally operational, I'll add some basic persistance for more testing. Dive in: procrastination is the disease of the failures.

Wednesday, September 21, 2005

C-Pound is Looking Sharp

I took the time to view the video interview about the LINQ project and C# 3.0's first-class query mechanisms. I was actually impressed with it, and it solved some problems I have with Python. It also made me think how the same solutions could be applied in Python, or any such dynamic language.


What is most interesting about LINQ, I couldn't be sure of from the video. Does it pull in all the data and apply the anonymous functions to it for conditional filtering, or does it take the where clause and actually generate that into a request to the server? Now, this doesn't matter for XML documents and especially direct queries on CRL objects, but with relational tables it really is important.


This is what Python needs: a generalized way to generate patterns to match, using existing operators. There are some tricks available, like allowing sometype==value to return a callable that can be passed an instance of sometype and return True if it equals value. This can be used to generate lots of pattern matchers. The problem is that some things aren't so easily redefined, like and and or operators, and it isn't easy to define that operation if you also want == for instances of sometype. That requires defining an instance method and class method of the same name for one type. Possible, but very tricky.


It would be very useful to add first-class pattern matching to Python. I'm thinking of a syntax along the lines of ?==x to generate a pattern object representing the pattern needed. What is important, however, is that it would create more than an object that can be called to test if something matches the pattern, but other code would be able to introspect the pattern, merge patterns together, generate SQL code from them, etc.


I wonder how well the community would respond to this. It would probably be best to promote a common and hopefully standard pattern package, and to propose syntax integration of the module at a later date. That would also give time for the pattern matchings to evolve. Hopefully this will happen, because Python is improving every day, but if it doesn't do so in the right places, it will fall from its rising seat of glory.


(More examples of the Pattern Syntax)
(All comparison operators work as expected, such as `?>10' to test for values greater than 10)

`?foo' to test for an instance of foo

`?is foo' to test is something is foo

`?in foo' to test is something is in foo

`?contains foo' to test if foo is something

`?hasattr foo' to test if something has the attribute 'foo'

`?attr foo ?==10' to test if something's 'foo' attribute is equal to 10

`bar?foo' to test if bar is of type foo, where bar is the name of an iterated object, such as in a for loop or list comprehension. This can be used to clarify when there are more than one object per loop to test.

`?[4]?=="four"' to test if the 4th item in a list-like object is equal to the string "four".

`?foo,bar' test if something is of type foo or type bar

`?all foo,bar' test if something is of both type foo and type bar


Any thoughts on this?