Re: bash logins: a little buggy (and potentially dangerous)?
Manoj Srivastava <srivasta@datasync.com> writes:
> Count me in amongst those who shall be up in arms.
You were top of my list of potential candidates :>
And in reality, I'd probably oppose this too if it were proposed as
new policy, but for many (most) users it would be nice to have some
way that they can take advantage of these things which are (arguably
AFAIK) better for the common case.
In the following I'm speaking about *most* people, not special cases,
and I admit that my unix experience isn't broad enough to be *really*
confident that I'm not missing important reasons why the following
might be bad ideas...
That said, wouldn't most people be better off if:
1) their .xsession were symlinked to their .xinitrc
2) for those who use bash, their .xinitrc sourced their .bashrc at
the top.
3) there were a straightforward place to put bash code that they
could be *sure* would be executed either:
1) only for top-level shells (for things that are exported to sub-shells)
2) on every shell invocation
3) in such a way that --login won't sneak in and undo their work.
Overall the learning curve for getting this stuff right for new Bash/X
users is fairly steep ATM, and some of it (at least the umask/--login
and PATH/--login stuff) is both undocumented, and a bit unexpected, at
least by me.
The changes I'm talking about could be argued as falling under the
"principle of least surprise" (though it *would* be surprising for
those who are used to the normal setup -- However, I'm not talking
about making this the default, so that's not relevant).
I imagine the only way to do this "right" is to just put up a web-page
or mini-HOWTO somewhere explaining what's going on, and proposing a
possible solution. I might do that if I get time. It would actually
save me some effort when I'm helping others get started with Linux...
--
Rob Browning <rlb@cs.utexas.edu> PGP=E80E0D04F521A094 532B97F5D64E3930
Reply to: