[openbeos] Re: The world as we know it

  • From: "Nathan Whitehorn" <nathan.whitehorn@xxxxxxxxxxx>
  • To: openbeos@xxxxxxxxxxxxx
  • Date: Sat, 21 Sep 2002 19:06:42 EDT (-0400)

> >What is also the case, though, is that we already *have* R5. It 
> > exists. 
> >I'm running it. If R5 is equivalent to R1, then we have R1. So why 
> > not 
> There is have, and there is have. :-)
> We have no source, no possible way to extend, to hope for the future 
> R5. Yup.
> Also, believe it or not, I believe that there is cruft in R5. From 
> Dr8,9 PR,DR3,DR4, etc.
> We will come through cruft free with something that we understand and 
> have the source
> to. Is that not enough?

There's *lots* of cruft in R5. But building R1 on R5, for that very 
reason, ensures that R1 will hardly be cruft-free. Further, you can 
extend it without having source. Look at the MDR, for instance. We had 
no source to anything at all. Yet we managed to extend the system, and 
replace part of it with something that is (IMHO) much better. You don't 
need to clone everything to make changes. That's the beauty of Be's 
modular architecture.

> >start building the cool things now? Sure, it would be nice to have 
> > it 
> >all open source. But we can, at the same time as we replace those 
> >components, make them better. There is no reason to simply go 
> > blindly 
> In a few cases we are. Where it is fall down, dead obvious. 
> mmap is a good example. So is networking in the kernel. How about 
> integrated
> FS/VM cache? All good stuff. But where we are drawing the line is in 
> creating 
> new APIs and changing the ones that exist. 

Of course, some of this (kernel networking comes to mind) is not "fall 
down, dead obvious". It's a design decision, and more deserve to have 
been made.

> *IF* there were some design out there that was legal to look at that 
> was R6, I would
> seriously consider it. If I could go into an alternate future and 
> come back with the BeBook
> from BeOS 2015, we would implement that. But, truth is, design is 
> hard. And designing
> a complete OS is very hard. And doing that as volunteers who have 
> never built an OS
> before? Well, I wouldn't say that it is impossible, but I think that 
> it is very hard. I don't know
> of any really successful attempts at that. Cloning R5's API is the 
> best way that I know to
> get going. I like the example of AROS - they spent years debating 
> design before they said
> forget this and built AmigaOS 3 (recreating the last release). Once 
> they did that, they knew
> how to procede from there.

I never suggested spending years debating design. I just suggested an 
evolutionary approach to upgrading the system, which is made quite 
possible by Be's architecture. This doesn't make blindly cloning R5 any 
less senseless.

Fortune Cookie Says:

Chicago law prohibits eating in a place that is on fire.

Other related posts: