From the Third Party Developer’s meeting we learn a bunch of inventory changes are ready to roll out to RC testing. But, they are waiting on the HTML HTTP* viewer, Azumarill. Once the viewer is out, soon, the HTML HTTP part of these changes can start rolling to the servers.
This is another significant change in technology. For users it will be mostly transparent. You aren’t going to have new features or buttons, at least as far as I know. You will have fewer problems and inventory operations will be more reliable and I suspect quicker. Less chance of losing inventory.
I’m not expecting this roll out will be easy. The viewer is changing away from UDP protocol communication for many things that use it now to HTML HTTP. The server side of the UDP channels will be going away over time, think 2016. So, third party viewers (TPV) have to make the change over too. My point being there is a need for coordination in several places.
Oz Linden says they had to change lots of the viewer code. The changes touch lots of places in the code. Various tools and features used the UDP code and they all have to change. Because TPV’s do more things than the Linden viewer, so they have more things that use UDP that have to change. So, merging in the HTML HTTP changes is a significant effort. That suggests this may take some time for TPV Dev’s to catch up. That also suggests the Linden roll out will move slowly and support new and old protocols for a time.
Once the HTML HTTP changes are at least in RC they will release another viewer version that builds in the Inventory Changes, which are being built on top of the HTML HTTP changes. This is a second significant code change and likely a complex merge for TPV Dev’s. So, we have 2 code merges coming from the Lab that will likely be a challenge for some TPV Dev’s.
Once the Lab starts turning off (removing) the old services and protocols server side, old viewers will break. Oz says that turning off UDP services is not in the plans for this year (2015). But, will likely be early in 2016. So, that puts it 3 to 6 months out.
The choice to break old unsupported viewers or remain compatible was made based on a desire to move ahead and fix inventory reliability. As it is now there are bugs that can cause inventory to move to unexpected places and for practical purpose be lost. There are also bugs that can cause actual inventory loss. To remain compatible with old viewers those bug could not be completely eliminated. So, old viewers will break and we get better inventory reliability. I for one would rather not lose inventory items.
*Due to a brain fart I wrote HTML when I should have been writing HTTP. The first is a language and the second is a protocol.
Surely you meant HTTP, not HTML?
You are right! I hate it when I do that…
‘The viewer is changing away from UDP protocol communication for many things that use it now to HTML.’
You are confusing HTML (Hyper Text Markup Language), which is used to describe a web page, and HTTP (Hyper Text Transfer Protocol), which is used, among other things, to transfer web sites HTML pages (but in SL’s inventory case, it’s LLSD-formatted data instead), itself being a TCP (Transmission Control Protocol, which got a built-in packet receiving acknowledgment, making it a reliable transmission protocol), as opposed to UDP (User Datagram Protocol, which doesn’t have a built-in packet receiving acknowledgement, the latter having to be implemented at the application level, such as in the SL viewer, if a reliable transmission is desired: that’s why you got a “packets loss” statistic in the viewer; it refers to UDP packets that got lost).
As for old viewers, as long as they implement the AIS3 inventory protocol (which relies exclusively on HTTP and is the only protocol that SL servers will support in the future), they will keep working just fine (the new RC viewers only remove old UDP paths, re-organize the viewer inventory sources tree, and make a large cleanup of the HTTP code in the viewer).
I not confused about them. I was having a brain fart… DoH!
This is pretty awful to read. For someone who has been reporting about SL viewers that long, you get a lot of things wrong. The worst mistake was probably confusing HTTP with HTML, although a webbased viewer might be interesting too. Maybe just a bad day 🙂
Your right. It was late and I didn’t see it when I proofed it.
HTTP
Right
Pingback: Second Life’s Mesh Mess | Nalates' Things & StuffNalates' Things & Stuff