User:SignpostMarv Martin/Archive/Implementing new features
Of course, a really valuable way to contribute is to add a new feature. Try to work with the community and with Linden Lab in planning the feature before running off and implementing new things. Though we appreciate your hard work, we can't accept every new feature, since maintaining new features comes with a cost. Try thinking of ways to use APIs to make plugins, or perhaps propose new APIs to make the viewer more extensible, before adding new things to the core viewer.
If you have an idea for a new feature you'd like to see in the client, consider adding it to this list.
Discussion on each topic can be found in this article's Talk page
Patching it so that llLoadURL opens the F1 Help Browser
Falk Bergman (by way of #opensl on IRC)
- Adding keyword replacement in llLoadURL, Streaming media support, and other external links
- For instance, to provide an icecast stream with a agent UUID, http://'''agentkey''':apassword@icecast.somehost.com:8000 as a parcel URL
- To provide an external website with a SL User Name, http://www.somesite.com/somephp.php?query=stuff&slname=''agentname''&action=buy&deliveryto=''agentkey''
Barriers to Implementation
- Unless the LSL2 VM starts supporting optional parameters, you risk breaking thousands of scripts
- Workaround: Prefix urls with ubrowser:, e.g. ubrowser:http://example.com, which will result in the dialog box indicating that the url will open inside the client.
- Alternative Workaround: Clientside rewriting only, for connecting to media URLs
Adding a Avatar Local Stream Channel
Falk Bergman (by way of #opensl on IRC)
- that has a Range of X Meters around an Avatar
- This would be used for Local Audio Streaming
- Or Local VoIP chat
- Teamspeak...
Peer to Peer Voice over IP using above idea, improved
- Capible clients advertise themselves via below CTCP-style protocol
- Clients use multicast packets to broadcast to a list of addresses the server manages of clients in range.
- In effect, this would give each avatar two unidirectional shoutcast-style stream impliments, or one bidirectional impliment.
Pseudo-Voice
Lee Ponzu 13:49, 12 January 2007 (PST)
Most of our computers have local, client-side voice synthesis. Yes, it sounds like a robot, but it is still voice. Real voice, ala Skype, is fine. But just like we use Avatars to hide our real selves, we need a way to hide our real voice.
A huge advantage is that we could get voice without adding to the bandwidth. The text is coming anyway.
- The simplest idea is that one could mark one or more Avatars, and their Chat would be rendered as voice, as well.
- Slightly fancier---One could set the voice characteristics (male, female, pitch, etc.) for each other Avatar.
- Fanciest---Each user can set ones voice characteristics using Appearance, and those setings would be used on each local client.
Barriers to Implementation
SignpostMarv Martin (15:46, 12 January 2007 (PST)):
Distributed Processing
- Processing the audio on the local client and streaming it in-world would require Peer-to-Peer streaming audio.
Local Processing
- Distributing text-to-speech preferences would require modifications to the server-side code, using 3rd party servers, or Peer-to-Peer protocols.
- Allowing for the ability to add processing the data for up to 400 avatars (100 avatars per sim, within shout distance of region corner) at once would add significant processing overhead.
While distributed processing would probably mean less CPU-intensive code, local processing would have the distinct advantage of allowing users' preferences to be overridden, allowing people to do such things as cap voice pitch, preventing annoying chipmunk-speak from occuring.
Addendum
Alderic LeShelle (00:55, 25 January 2007 (PST)):
- It is possible to restrict changes to client side only. One may use existing chat/IM protocols to transmit speech preference data or use a client-to-client protocol to achieve it.
Distributed processing would mean a considerable network bandwidth impact, be it on each client or a dedicated chat server where local processing would simply add the chat preference data traffic to the network, but the speech synthesis CPU load on each client. It is easier to cap the CPU load, though, for example by simply capping speech synthesis for avatars out of a certain, configurable, hearing range, or simply disabling speech synthesis for the nth nearest avatar and further away.
More efficient local cache
I'm working on this. We are going to use Berkeley DB with ARC replacement policy. It's going to fly. Gigs Taggart 17:54, 31 January 2007 (PST)
CTCP protocol layered on IM system
Barriers to Implementation
Establishing a circuit between two clients will probably require server side assistance, since circuits are UDP based, and most clients will be behind NAT's and firewalls. CTCP will require a clientside listening socket either implemented via RFC 3489 or some sort of UPnP code using a TCP direct connection.
Command Line Interface Improvements
- for changing preferences sending an IM, teleporting, etc
- for example: "/set drawdist 96", "/set sound off", "/tp ahern", or "/im kex godel hello"
- would need an escape system which doesn't conflict with script command gestures (or a reserved words like how /me is)
- Or better yet, the Ctrl-G Gesture management window could be improved to allow this, and even allow rebinding /commands
Barriers to Implementation
- Implementing an escape system which doesn't conflict with script command gestures (or a reserved words like how /me is)
More sophisticated IM features
- muting/filtering/autoresponse options
- mute or do not alert on IMs by started by a group, agent, or group-agent
- Autoreply to IMs received while (Away)
- Sidebar Userlist of Active Users who are currently participating in a Group IM Session.
- When a user says something, they're added. When a user leaves the session, they're removed from the list.
- More control over filtering specific objects/textures/sounds/agents/etc
Barriers to Implementation
- Anything that requires server-side modification SignpostMarv Martin 11:11, 11 January 2007 (PST)
Shortcut/Link/Alias function in Inventory
- Add a "link" ("alias", "shortcut") facility to the inventory, that would not break Permissions. For instance, I have one nice set of prim dress shoes that are no-copy, and four tuxedos. I cannot make complete separate outfits, because I have only one copy of the shoes. However, using a "link", I could create an outfit folder with the tux and a link to the no-copy shoes. Still means I would only have the one set, but it would make changing clothes much much easier. Buckaroo Mu 09:23, 9 January 2007 (PST)
DirectX3D Hardware Acceleration
- A DirectX3D-driven version of the viewer, allowing those with ATi or Intel hardware accelerated cards to take advantage of the processing power. The system could allow selection as a preference of OpenGL or DirectX3D, as many Windows-only games currently do. Conditional build statements would keep the option from appearing in the non-DirectX3D capable builds. Buckaroo Mu 09:28, 9 January 2007 (PST)
Barriers to Implementation
- Developer incentive/motivation
GPGPU Support
- A method of offloading some of the graphics processing to the GPU might help increase framerate. Buckaroo Mu 09:37, 9 January 2007 (PST)
Multiple Monitor Support
- Support for detachable sub-windows, allowing the inventory, chat history, IM, etc. windows to be moved to a second monitor (if available). This is a very highly-desired feature (by those that have dual-monitor systems). Buckaroo Mu 09:37, 9 January 2007 (PST)
- This could potentially be done using an external program, if SL had some way of exposing this data in a crossplatform way. (Does NT support named pipes?) Some libsecondlife projects such as SLeek are already exposing this. Kamilion Schnook 20:39 Jan 9 2007
Implement GUI in Native Widgets
I also would like to see multiple monitor support, but I'd prefer this to happen by implementing the inventory system and other parts using native palettes, toolbars, and windows (dockable to the SL window).
Embedded scripting language for client-side plugins
Heather Goodliffe, Yumi Murakami
- Make the client more powerful and plugable
- Embed a scripting language, such as LUA or Python, within Second Life to enable client plug-ins to be written in a modular fashion. Client plugins would be distributed via Second Life itself; hopefully LL would eventually agree to create a new object type for these, but for testing purposes notecards would probably suffice.
- Let me second that: I'd like to see client-side scripting as well. It should be easy to add that (in particular, using Lua), and it would let users fix so many usability problems. At a minimum, the scripting language should be able to access the functionality available through the menus, it should permit key bindings and grab keys, access preference settings, and it should be able to listen to chat and IM, so that commands can be triggered from there. Jherek Cerminara
Nude patch
- Modify the client to not render any avatar clothes or attachments.
Barriers to Implementation
Steps to how Linden Lab could 'disable' a 'nude patch'
- Run a check on clothing textures including alpha information
- Check for alpha information around the genitalia, refuse upload if check returns positive
- Composite clothing textures server-side when avatars are:
- In PG regions
- Viewable from PG regions
A 'nude patch' would function by:
- somehow submitting full alpha textures on all clothes no matter what was worn (prevented by LL as described above)
- ignoring what the server says regarding skins
- Could use procedurally generated textures based on avatar shape, and primitar type to prevent 'Attack of the (nude) Clones'
- SignpostMarv Martin 01:02, 11 January 2007 (PST)
- I don't think a nude patch could work anyway - the skin and clothing textures are combined on the wearer's client, then the combined texture is uploaded to the server as the single texture that appears on the avatar. That's why "rebaking" your textures on your client affects how other people see you. You could possibly just apply the same skin to everyone but you might just as well leave everyone unrezzed then ;) - Yumi Murakami
- here is a quick and dirty solution ;)
Two Tier Search
- Server performs a possibly less-effective search which is database efficient, such as the current fulltext search. The client then performs its own search, refining the results. Should allow extra flexibility and take some load off LL's database.
User Interface Improvements
- General improvements to the user interface. Maybe move to a seperate article: User Interface Improvements
More Text for Abuse and Bug Reports
- Bug Reports as well as Abuse Reports are currently very limited in the amount to text you can enter. Very difficult sometimes to describe a bug in just 2 or 3 sentences...
NOTE: A lengthy bug report that would benefit from additional material such as videos or photos could possibly be submitted to Jira --SignpostMarv Martin
Attachments registering as additional limbs for animation
- Attachments with distinct object names (like wings, tails, extra limbs...) may register themselves in the skeletal structure of the avatar, so that a BVH animation might animate them as well.
Much Longer Land Draw Distance
- The idea is to have a separate slider for land draw distance. It must be greater than or equal to the "master" draw distance. As drawing land is much less resource-intensive than drawing objects and avatars, it should be possible to have a land draw distance measured in kilometers, not meters.
- Could possibly use colored map textures (from the world map, not from the viewer-generated local map) as terrain patch, rather than the supplied patch, although this might not look great for closer-in terrain.
- Perhaps all terrain within 512 meters could be drawn using the sim-supplied textures, and anything further could be drawn using the world map textures.
Barriers to Implementation
- Would require child connections to more simulators to render "sorta near" terrain (that is, beyond the draw distance for objects and avatars, but within the usual "far clip.") The simulators would have to JUST send land, not objects/avs/etc. as that would defeat the purpose of having a lower draw distance. (Terrain beyond the usual far clip would all be done as world map textures.)
- Colored map textures may not be available in suitable resolution for it to look really good. However, it should be acceptable to many residents in lieu of the "version 2" graphics engine we were shown previews of in 2005 (and which subsequently was cancelled.)
- If this works well as a proof of concept, LL might be persuaded to make higher-resolution world map textures available, and work on some server-side optimizations.
See Also
Linden Lab is considering offering bounties for especially desirable features in the viewer.