Anybody working on removing sendmail from base?

Matthew Dillon dillon at
Thu Oct 2 10:32:46 PDT 2003

:the downside I see to useing kernel memory is hte question of just how much 
:memory the structures will take. In a system where 100 or 1000 processes 
:could be running at once, if each of those has a varsym table for it, and 
:there are 3 versions of perl and 3 versions of gcc installed, we are going to 
:be using a good chunk of  memory.  even if each layer only stores 'overrides' 
:from the previous layer, there may only be 5 users, and there would only be 
:one system are still talking a decent chunk of memory real estate.

    I don't think this is an issue, there is no reason why we would actually
    have to duplicate the syms for each process if they haven't changed.

:Also, isn't the idea of variant symlinks to make things dependent on the 
:environment?  It seems the idea you have discussed makes for having to use a 
:different way to set such things.  Question: can't we simply parse the 
:environment from namei() if a varsym is found, or would this incurr too great 
:a performance hit? (in other words, if we find a ${mta} we search the 
:enviroment for $MTA, which would have been set globally by /etc/rc.d/mail (or 
:overriden by .login or something), but if we don't find a ${ in the path, we 
:treat it normally?  Or am I missing something (which should be) obvious?

    The environment is in user space, which means that namei/lookup would
    have to go accessing user memory, and the environment is also not
    indexed so lookups would be slow.  And, even worse, the environment
    *IS* duplicated in each process.

					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>

More information about the Kernel mailing list