[sldev] Re: [META] Improving the source code

[Lawson English]

Nicholaz Beresford wrote:
>
> Michael Miller wrote:
>> Let's revamp the client, and show them that we can do
>> something useful!
>
> I can't help it but this reasoning sounds pretty backward.
>
>
Unless you're hoping to be hired by Linden Labs of course... ;-)

Parts of the code, such as the GUI, ARE being redesigned internally. 
Benjamin Linden has helped set this agenda, with input from many of us. 
There are still at least a few things missing from the user interface 
roadmap, though, which is where most of my meager abilities lie.

I am sure there are many changes planned on the level of the client 
architecture that haven't, publicly at least, been vetted by the open 
source community as thoroughly as the user interface stuff has been. It 
might be good to start drawing up nice little boxes with arrows showing 
data flow, class hierarchies, and so on, both for the existing client (a 
spaghetti mess will show up, I suspect) and for a hypothetical 
redesigned client. Replacing one threading library with another might 
tidy up things a bit, but I suspect that the REAL design flaws are at a 
more fundamental level (given what I have seen of the GUI framework) and 
that any replacement of threading, etc, libraries, unless it is a 
trivial task to complete, should be made during the implementation of 
the  redesigned  client, rather than piecemeal.


Case in point being the GUI framework, which is hopelessly entangled, as 
is. The UI roadmap calls for sufficient refactoring to eventually 
provide a viewerless GUI test-client AND a GUI-less client library that 
can be ported to other systems in a form other than as a standalone 
application in home workstations (Second Life on an iPhone III?) This is 
a double-plus good thing, and is essential for any REAL redesign of both 
the viewer AND the GUI framework to take place.

But that doesn't mean that we can't start discussing these issues and 
drawing pretty boxes and sharing them now, rather than waiting until the 
initial divorce of the GUI and viewer is complete. Certainly, I'll be 
buttonholing Benjamin regularly with my pie-in-the-sky user interface 
ideas (like optionally linking WindLight themes to GUI skins when you 
enter a new sim,  for example, with HUD controls that modify the 
appearance of the GUI clientside. Howabout combat sims, where the 
explosions not only light up the sky, but also are reflected on the 
pitted-metal, military-HUD-like client GUI. Or the Goth Masquerade sim 
that does the same with lightning and creepy-looking GUI textures? ).

L.