| Rika | 1. No, no textures whatsoever are loaded by HTTP other than map textures. Period. I studied it with a packet sniffer. |
| Rika | 2. Yes, textures do load faster otherwise anyway. That they load slower for others at the same time implies that they did something to UDP |
| Rika | fetch too, and not just textures. |
| Rika | 3. As usual with official viewers, an NVidia GX260 and Core 2 Quad 'do not meet minimum requirements'. |
| Rika | 4. Some textures - the ones that load faster - fail to load completely at all. |
| Rika | 5. First it cleared my cache. Then it put it's own into a different location /anyway/, bloody thing. |
| Rika | In short, not fit to be a main viewer yet, though a diff of what exactly happened to UDP fetch would be interesting. |
| Rika | But my guess is, it's a different packet size or something simple... |
| ChefKarl | interesting analysis. |
| Rika | Upon some further reflection and waiting for inventory to reload with an open texture console... |
| Rika | 1. It also seems to load LESS textures. I.e. the number of simultaneous fetches is smaller. |
| Rika | 2. Right now I'm staring at the same texture being reloaded from the sim every second... |
| Nikadreamscape | I got 3 blue screens of death in a row, using it. |
| Rika | If it actually limits textures to load only to those that can be seen, most texture prefetching will become useless. |
| Rika | The viewer side of the HTTP texture download code is completed though, no doubt about it. If it worked, it would get rid of 80% of all lag. |
| Rika | It's just that the server side doesn't exist. |
| ChefKarl | wonders | if and when they will get the server side finished to help reduce lag? |
| Rika | There has been no word whatsoever on the topic that I heard. |
| Rika | The code is pretty clear - viewer asks the caps system which URL to get textures from ("GetTexture" cap request)... |
| Rika | ...and then http-requests them off there giving texture key as a GET parameter to a web-something (which can be anything). |
| Rika | That anything doesn't exist yet, so "GetTexture" request results in nothing. |
| Rika | Mind you, they had the viewer side, unfinished and commented out in the published source, all the way since when it was first published. |
| Rika | Well, it's now finished. It hasn't even been three years. |
| Rika | How time flies </sarcasm> |
| ℧cciekins | says | Thanks! Didn't understand a lot of that but what I did explains much of what I've seen and sounds really cool. |
| Vεχtrα | Rika, I noticed an advanced menu option to turn on HTTP textures.. I'm sure you would have toggled this on in your tests? |
| Sylvie | says | I've been using it. Been getting a strange error where prims rezzed onto a surface (eg table) sink slightly into the surface |
| Sylvie | says | Not sure if that's snowglobe or my prims... |
| Sylvie | says | sorry, error is the wrong word. Strange *behaviour* |
| Rika | Yes, obviously I turned the option on. I had it on forever, actually, because internally it's still the same ImagePipelineUseHTTP |
| Rika | Prims rezzed onto a surface sinking into that surface happens to me in 1.22 sometimes too. |
| Rika | I think it's related to floating point and altitude. |
| Peter | says | blog the analysis |
| GwynethLlewelyn | says | Thanks, Rika! Soft Linden tends to confirm what you say, too, although nobody at the lab seems to be aware of what Philip's group doing. |
| GwynethLlewelyn | says | They all believe that nothing works on the server side yet. Hmm. I wonder, though, if we could get OpenSim to work with it |
| Rika | Get it to answer caps request with a sensible root URL and put a php script nearby that extracts blobs from sql asset storage, it will. |
© copyright 2007-2012 Plurk Search People Search Verified Account APIFAQ
Contact Us About Blog Jobs Terms Privacy
