Re: rc improvements (wanted?)

There is a list specifically for discussing rc-related stuff, 
freebsd-rc_at_freebsd.org, FYI. I have a lot of questions which are not 
intended to be critical in anyway, it's just important to think about 
stuff carefully before making changes to the boot stuff. You might 
want to reformulate a post that has the answers to the questions below 
and start again on the -rc list, but that's your choice.

David Naylor wrote:
> Hi,
> 
> A while back I worked on an "improvement" for rc.  Most of the work was in 
> recoding rcorder.  
> 
> The 'improvements' in rcorder:
> * Add -e -i commands (allows simplification of rc by removing need for 
> early_late checks)

Can you explain how you accomplished this?

> * Add stagnation or parallel support (all scripts in a stage can be executed 
> concurrently without conflict)

How are you defining stages? Is this the "minimal change in the rc 
scripts" you're referring to below?

> * Marginal speed increase (irrelevant since previous version is fast enough 
> [unless one is worried about milliseconds on start-up time])

Faster is better.

> The main reason for this work was to increase start-up time (on the userland 
> side) by running as many scripts concurrently as possible. 

How are you running the scripts concurrently, and the key question, 
have you actually benchmarked your changes to demonstrate that they 
result in statistically significant changes.

> This approach 
> allows only a minimal change in the rc scripts (there is a more efficient 
> method but that would mean moving most of the controlling logic into a 
> binary).  

When you say "controlling logic" are you referring to what /etc/rc 
does currently? Replacing rc with a binary is not out of the question, 
but we'd need pretty clear evidence that it's the right thing to do 
first.

> I am eager to continue with developing the above if the FreeBSD project (and 
> developers) want such a change?  Or alternatively I could pass on the work 
> already done to someone interested.  

Posting a URL where people could examine your work would be useful.

> [[Side note: I stopped  short of actually field testing the concurrent changes 
> to rc (rcorder and the simplifications to rc scripts works]]

Ok, that answers one of my questions. The last question I had off the 
top of my head is whether or not you've tested all this stuff on 
shutdown too.


hth,

Doug

-- 

     This .signature sanitized for your protection