Allowing Remote HTTP Communication

Adding the ability to do remote HTTP requests would help out a lot in scripting projects. Currently its extremely difficult to store information in SL. A lot of people have expressed interest in being able to set up their own notecards to store information. I can understand that this could put a large strain on the servers if everyone was storing info in a flat text file. How about letting us deal with the extra CPU load? I know that myself and some other residents have access to servers with web and databases running on them. If we could issue HTTP requests (and recieve the responses) I'd even help set some people up on my own machine.

I can understand the potential problems with this situation. Most notably would be the extra traffic and possible illicit uses. I have two suggestions for dealing with these issues.

First, put a maximum request/response size on all HTTP communication. In this way people won't be exchaning mp3's using this feature. It will also make sure that an excess amount of load is not put onto the servers parsing the streams.

To thwart the problem of misuse directed toward foreign servers I have a few suggestions of varying degrees of effectiveness. The most restrictive option would be to have residents register a list of servers (based on IP or domain) that their objects could access. This would make it difficult for some members to use this feature, but would eliminate most of the potential to hack CNN from SL. You could also impose a fee on the external requests in L$. I would think that this fee should be quite small (on the order of 1L$/100 requests or smaller) to enable residents to fully enjoy this feature.

For the initial release of this feature, HTTP GET support would be sufficient. One neat feature I think people would use right away would be the ability to control a Shoutcast stream from inside SL. There has already been some implementation of SL-run streams, now they could control it from inside SL! I think that allowing limited HTTP capabilities would allow much more flexibility in scripting without adding much additional stress to the servers.

Besides the possible naughty-ness that may arise, I woud be in fear of conflicting protocalls and the lag that it would cause. Maybe that's just me.

I hear you.

Such a system can be implemented in a way similar to the dataserver requests. Do a llRequestURL( string ) call and when the response comes back, a web_result( string ) event is triggered with the data. Perhaps limit the data size to 1k or, if ya gotta, 256 characters. Strip unicode out if SL doesn't support it. (Don't know if it does or not.)