I wanted to follow up on the platform discussion a bit. No question it is a 
very large issue that looms ahead of us, and the points discussed earlier are 
very well taken. One of the main reasons I like the PowerPC arch is it's 
versatility, integration and wide app range. But with this flexibility comes 
non standardization and complication in the whole target discussion...  but I 
want to speak tactically for a moment. 

The ppc-dev project's base goal has always been to get a  kernel up and 
working. We are very close to having a single user prompt (SUP) up on the ODW. 
This has been alot of work and even if we get to the SUP we have tons more 
development to do before it is ready for primetime usage . This is work that we 
hope others now can jump in on. We have moved everything out to the project 
site and are doing commits there bi weekly. We continue to upgrade the build 
tools and the documentation. This development is happening here.

It is also why we are now working towards the shared target approach. We will 
have shared build servers wih multiple ODWs, available externally such that 
involved developers can log into and run their code. We will find enough ODW 
targets to satisfy this need. The approach has worked well for us internally 
and should work well for anyone serious about development here going forward. 

Yes, eventually we need a longer term target solution but what that ends up 
being exactly for this community will be driven, IMHO, by factors related to 
the business adoption of solaris on ppc. Until that happens I cannot see us 
losing focus on our immediate goal of completing a working kernel, even if that 
is on a target that is not readily available to the masses.

I also want to mention that we haven't spent a great deal of time refining 
specific platform work here. Since the ODW was very "PC like" there were a 
number of x86 based approaches that were straightforward to follow and utilize. 
We however have given a good deal of thought along the way of how to architect 
areas of solaris that would lower the porting effort on future targets.  

Additionally if we ever get to the point of needing targets for contributing 
developers and can't find additional ODWs, we do have access to a number of ODW 
like targets that we had developed in the Labs a while ago. Anyway, I'm 
confident we can continue to make progress with what we have and limit the 
amount of throw away code we develop on the ODW. But forward progress, and 
getting more of solaris up on this architecture is what we must keep as a 
priority.

Tom
 
 
This message posted from opensolaris.org

Reply via email to