|
View previous topic :: View next topic
|
| Author |
Message |
casey
Site Admin
Joined: 18 Dec 2004
Posts: 1768
Location: Seattle
|
| |
 Posted: Tue Nov 17, 2009 11:54 am Post subject: Re: Jeff and Casey Podcast - Season 2 Episode 26 |
 |
|
| rotissarychicken wrote: | | 13 and the jew are finally back on the team and like fUCK YES |
I've already voiced my opposition to the 13 idolatry, but I will mention that I am surprised no one has coined the phrase "13 is the new 10". Or something.
- Casey |
|
| Back to top |
|
 |
rotissarychicken
Joined: 30 Sep 2009
Posts: 92
|
| |
Posted: Tue Nov 17, 2009 10:14 pm Post subject: |
|
|
hey dudes 13 is the new 10
also what does idolatry mean
 |
|
| Back to top |
|
|
rotissarychicken
Joined: 30 Sep 2009
Posts: 92
|
| |
Posted: Wed Nov 18, 2009 12:38 am Post subject: |
|
|
| ryg wrote: | Wow, lots of interesting stuff in there, and plenty of stuff I'd like to comment on.
First, the whole demoscene thing. I'm basically with Jeff on this one - it's definitely the results that matter; if it looks crappy then I'm not interested no matter what. Besides, making something really tiny isn't nearly as hard as a lot of people think it is. Don't underestimate the selection bias - you only ever get to see the 256-byte (4k, 64k, ...) intros that actually got released. You don't get to see the stuff that the coders never managed to squeeze in less than 280 bytes, nor do you see the versions that were 210 bytes and a lot crappier. The last part especially - of course all the really tiny stuff is packed to the brim, there's no point in not doing it, and filling 16 bytes with code doesn't exactly take long no matter what you do 
Anyway, most 4k intros nowadays are written entirely in C/C++ (Crinkler's doing - so much for the claim that it's all hardcore Assembler tweaking), and at least half of all 4ks released last year mainly contain code for sound synthesis plus a single GLSL fragment shader that raytraces/raymarches to render. It's basically the perfect playground for people that are really good at picking just the right magic formulas (and magic values) to make something look good, and the demoscene is certainly a good place to find them.
These people do have a sense of what's cheap and what's not, so they're certainly far more grounded than most programmers are nowadays, but the ratio of actual systems programmer types in the demoscene is definitely not as big as it once was, and it's steadily falling. There's several reasons for that; for one, there's really good tools even for demoscene-exclusive purposes like extreme size optimization by now. Mentor and Blueberry wrote Crinkler for 4ks, and I did kkrunchy for 64ks, and between the two there's little sense for other demoscene guys to even write their own packers anymore, since that problem is pretty much solved by now - both of them are within a few percent of the best-known packers for stuff in their size region, with quite small depacker stubs. There used to be dozens of people writing packers for demoscene stuff, now it's really just Mentor and Blueberry. Same goes for a lot of other things; music used to be made with "trackers", and there were lots of people writing them, and lots of people writing players. There's been several player libs that *just work* since the late 90s, and nowadays everyone just uses MP3/OGG anyway, so the tracked formats are useless anyway. There were dozens of paint programs; nowadays everyone uses Photoshop, and even though PS sucks hard in a lot of respects, the entry barrier is just far too high for one or two programmers to write a serious alternative in their spare time as a side project. The list goes on.
Basically there used to be a whole "ecosystem" (I like to make fun of that word when applied to software, but it fits here) around the demoscene which is what drew a lot of the systems programming people there, and that's just gone now. If you're a systems programmer in the scene now, you're either one of the very few (maybe 15-20) who actually write tools that someone else uses, or you write demosystems. Theoretically there are also guys that write engines and content pipeline stuff; but in reality the coders that do this all work at game companies and do the exact same thing there too. All this shit is just too horribly broken to figure it out in your free time by yourself "for fun", unless you're actually insane.
We (farbrausch) started doing our 64k intro thing in 2000, and what we did was really mostly systems work. Our group is different from most others in that we almost exclusively have system programmers, and not one programmer that's really good at finding magic values to make some effect look good. 64k is actually a pretty nice size - you can really do something decent, and you don't have to tweak the implementation all the time, but you really have to get the design down and minimal. More importantly, it really is about design at that size; less and it's a lot about micro-optimizations and ends up very hacky. We did a lot of interesting stuff, not necessarily in the results, but certainly in the way we got there - and we really iterated a lot on this, making the core components simpler and more orthogonal every time. It definitely improved by coding faster than anything else I ever did before or afterwards - can't speak for the others. But we basically ran out of interesting challenges for 64ks by 2003, then we did kkrieger (released in 2004, and really just a tech demo for a tech that nobody needed or needs), and then debris in 2007. The latter was already "okay, stop the fixed size limits, let's just see how far we can push this approach". And it was fun, but now it's over, and anything we've tried to do (and done) afterwards severely suffers from none of us really wanting to deal with the engine/content pipeline problem.
Which ties in with your bit about graphics programming in general - and it really felt good to hear that I'm not the only one thinking this. I never liked hack-heavy programming to begin with (there are lots in the demoscene who don't mind, but I'm not one of them), and the whole lengthy size-optimization stint made me completely hate it. And that really ruined graphics for me during the past few years. The state of the art really is a combination of ridiculously fragile techniques with huge amount of manual (artist) labor to hide all the problems. And I hugely respect Crytek etc. for actually managing to get anything done with these huge teams of programmers that implement special-case hacks, and even larger teams of artists to produce that content, and probably a still larger QA department to find all the problems before release. But I can't respect or even like this "solution", because that's not a solution at all, they've just given up. And I may think about doing things this way at work if I have to, but there's no way in hell I'm gonna spend my spare time on something that's this ugly.
Actually, that last thing has been a serious problem for me lately. I do have a relatively big set of interests coding-wise, but basically everywhere in computing, things escalated into this arms race of piling more and more ideas onto each other for ever smaller gains, losing all conceptual unity in the process. And it's making everything worse on all levels. In the bad cases, you have software that's just layered to death and nobody knows what's really going on anymore - the way desktop OSes and browsers have been going, all of them. And the best-case scenario is software that's still functional, but tweaked to hell and back, with tons of special cases everywhere. Even if you have a better design, you still look worse, because the other code has all these special cases worked out and you haven't, and it creates this tremendous inertia that's completely stifling innovation. It's killing me. |
dude you type too much
also you guys should have a house segment on the podcast, that would be like totally rad. like, 10 minutes (1/6th bitches) of the the podcast would be a review and shit of the last episode k.
it would be called, like, heil house (because cause house is awesome and heil is german and german is awesome) and it would be a about the last house episode and a review of the that last episode.
actually fuck that, you guys should have like a serparate podcast feed called heil house (as explained before) and itll be 30 minutes long of minutes you talking around the subject of the house epsiode.
you shall celebrate with bread and wine and thine ale with thine sandwhich
also do you dudes watch curb your enthusiasm? that jew dude is hilarious with the jew comedy and hbo shit. hbo is a rad channel because theres no commercials and most of the shit there is usually up on pirtatebay the next day |
|
| Back to top |
|
|
rotissarychicken
Joined: 30 Sep 2009
Posts: 92
|
| |
Posted: Wed Nov 18, 2009 3:23 am Post subject: |
|
|
yo guys i get my own thread lol
heres link to torrent of rad house epusode |
|
| Back to top |
|
|
rotissarychicken
Joined: 30 Sep 2009
Posts: 92
|
| |
Posted: Wed Nov 18, 2009 4:42 am Post subject: |
|
|
yo did the torrent link get removed?
i thought you dudes were all pro piracy and shit. anyway http://www.mininova.org/tor/3158759 |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
FAQ
Search
Usergroups
Log in
Register
