Re: Bootstrapping

[John Sechrest]

"Luke A. Kanies" <luke@madstop.com> writes:


 % >  Yes, but to do this, the cfengine community needs to start abstracting
 % >  the (global practices) from the (local issues).

 % That's a really good idea, but I'm not totally convinced this will be very
 % easy with cfengine.  Mark and I have had long discussions about this, but
 % my opinion has not changed:  It is very difficult to use cfengine for
 % abstraction, which means it's very difficult to create
 % location-independent cfengine configurations.

 I am finding that if I can seperate the "rules" from the "data"
 about which domain it is or which service I want to run,
 Then I can get pretty close.

 That is why I have been looking at reading the data that is 
 normally written as constants/Classes in the Cfengine
 rules into datafiles as a base experiment.

 And it seems to be working fine. 

 I can specify in a file what the "role" of a system is, and it 
 works to do it. 

 And I can move that role around, without rebuilding the machines.

 The simple experiments work.

 So if we can articulate the rules, and abstract the data in those,
 then I think it is possible to use cfengine to represent
 the details of a set of rules. 



 % >  A) Why would this not work as a component of the CFengine work that
 % >     is already going on?

 % Because it's essentially separate from cfengine itself.  It's a lot more
 % than just cfengine.

   Right now, I am writing many little tiny cfengine scripts. 
   Other than the infrastructure to name the roles and set the 
   base system, everything else seems to be cfengine. 

   So It seems like it could fit into the cfengine discussion 
   easily. 



 % >  B) Why would it need to be a seperate project?

 % Because it involves about 15 other packages and all of their
 % configurations.

 Interesting. Can you help me understand those other packages and
 configurations you are working with? 


 % >  C) Do you care that there already is an ongoing project working
 % >     on the same issue? (We got our sourceforge project approved recently)

 % Heh. :)  When it rains it pours.
 % I'm certainly willing to work with others, and I bet Nate is also.

 I am glad to hear that. Right now, I think one of the good things
 to work on would be to abstract the definitions of roles
 and what those mean. 

 Our UML configuration tool (MLN) is proving to be valuable in
 my process of doing that. I hope we will get the source forge
 site up with the .7 version soon. 

 Can you help us take a role of that list that went around
 and convert it into an abstract set of rules 
 in english (or at least predicate calculus), so that we can 
 talk about it before we implement it in cfengine?




-----
John Sechrest          .         Helping people use
                        .           computers and the Internet
                          .            more effectively
                             .                      
                                 .       Internet: sechrest@peak.org
                                      .   
                                              . http://www.peak.org/~sechrest