Package systems and stuff

Kendrick kendrick at
Wed May 12 03:43:05 PDT 2004

Robert Connolly wrote:

>On May 11, 2004 08:36 pm, Archaic wrote:
>>On Tue, May 11, 2004 at 05:23:33PM -0400, Robert Connolly wrote:
>>>What do you guys think about using a bsd-like source tree (using
>>>/sources)? I
>>Seems a bit too out there. Takes away our focus, IMO.
>I thought of this because we don't need a system that can make binary 
>packages, just a way of being organized. Its not just for vanity. Many of the 
>packages can build in an seperate directory, for a read-only source tree, but 
>there's an interesting problem with gcc & perl since it uses different 
>patches from chap5 to 6. Cross building would be more managable, or at least 
>this would give a framework for it. The permissions and ownerships of the 
>source would be better controlled; a package builder user would only need 
>write permission to the object directory and not the source itself, and then 
>only needs group read so (modified) source isn't world readable. A small file 
>could have the patch name and commands used to make /sources/bin/bash/ so 
>diehard users could still make their own tree. At least from my perspective 
>importing the core into one tree is easier to manage. It will also make it 
>easier to manage glibc and 2.6 headers in the future.
this sounds interesting i have no vote either way it might be a wise 
thing to do further down the road though.  might want to get v1.0 out 
then set this up as we go from there.  the tree is a nice feature  how 
ever i do believe it is to early to get started with that yet.  (maybe 
around time for getting X included?)

btw the automated  script is still a good thing.  how else is a sys adm 
to be able to quickly roll out a great deal of new machines in a 
reasonable amount of time else wise ;)   once there is a "simple" method 
of management  it brings a "dumber" class but also brings a larger 
audiance.  I have been reading through the man's for alfs and am not 
shure if it would be posible to get the hint info together for a 
standard user so they could substitute it for the book.  how ever i do 
know that there is great posibility for using it in a mass production 
environment.  ie  geting 3 servers and 25 workstations out in a 
reasonable amount of time.  a tool like that could help sys admins who 
understand the value of this project better approach exec's with the 
idea of using lfs style systems :)    it could be our  rebuttle against 
the dreaded windows installer.  we CAN easily make a initial system 
secure and with only what is needed for the system.  </strange rant like 

just my 2 cents

More information about the hlfs-dev mailing list