[sldev] back to the lab again?

[Michael Schlenker]

Tigro Spottystripes schrieb:
> Reading somthing on the forum inspired write this message, what are your
> opinions (both LL employees and community developers) about starting
> over  avoiding mistakes made with the first attempt of making SL , many
> that got discovered only too late and now piled up and got to be
> mantained, and go more carefully this time, taking input from users from
> the start so there is less chances that feature suggestions conflict
> with intended behavior and "undocumented features", already from the
> start only using code that won't have licensing concerns regarding
> opensource etc, not putting any effort in making this new SL/compatible
> with anything in the first incarnation, no content to be broken 'cause
> there is no content made yet, 
Here it breaks down. If you kill off all or even most of the current
content you could just start a totally different thing. Technically
nice, but commercial suicide.

> no bug exploits that gotta be supported as
> if they were features etc. Basicly, what are your opinions regarding
> getting a clean slate and this time doing things with much more
> planning, community input and testing etc?
Its the same wishful thinking you always find if you have a messy state
and need to clean up. Its well discussed, one classic piece about it is
for example,
http://www.joelonsoftware.com/articles/fog0000000069.html
there are others advocating a rewrite, but most people tell you its a
not so great idea as it sounds usually.

There are a lot of issues, sure, but you nearly always fare better if
you improve things step by step instead of doing a full rewrite,
especially if you have a big system. Getting a good test harness is
usually a big win in such cases as you know you don't break too much
stuff while refactoring and improving.

You can deal with all the issues incrementally, for example if you see a
bug gets feature status either develop it into a full feature and
deprecate the old behaviour over the next few versions. Similar things
happen all the time, see http-texture branch, internal reorg of servers
and queries, etc. etc.

Michael