Instead of `risky-local-variable'

Instead of `risky-local-variable'

[Richard M. Stallman]

I'm considering a more reliable replacement for the
`risky-local-variable' mechanism.  My idea is that only defcustom'ed
variables could be specified in a local variables list, and that the
:type together with the specified value would determine whether a
variable is safe.

What do people think?

Re: Instead of `risky-local-variable'

[Luc Teirlinck]

Richard Stallman wrote:

   I'm considering a more reliable replacement for the
   `risky-local-variable' mechanism.  My idea is that only defcustom'ed
   variables could be specified in a local variables list, and that the
   :type together with the specified value would determine whether a
   variable is safe.

   What do people think?

I personally think that this sounds like a pretty radical change,
supposedly "relatively shortly" before a release.  Any reason why this
can not wait till after the release?

Sincerely,

Luc.

Re: Instead of `risky-local-variable'

[Chong Yidong]

> I'm considering a more reliable replacement for the
> `risky-local-variable' mechanism.  My idea is that only defcustom'ed
> variables could be specified in a local variables list, and that the
> :type together with the specified value would determine whether a
> variable is safe.

This won't work for defcustoms that haven't been loaded yet.

Re: Instead of `risky-local-variable'

[Richard M. Stallman]

    > I'm considering a more reliable replacement for the
    > `risky-local-variable' mechanism.  My idea is that only defcustom'ed
    > variables could be specified in a local variables list, and that the
    > :type together with the specified value would determine whether a
    > variable is safe.

    This won't work for defcustoms that haven't been loaded yet.

It would probably refuse to bind the variable in such a case, but is
that really a problem in practice?  A file local binding is useful
because the variable is going to do something, which means the package
that obeys it needs to be loaded.

Re: Instead of `risky-local-variable'

[Chong Yidong]

"Richard M. Stallman" <rms@gnu.org> writes:

>     > I'm considering a more reliable replacement for the
>     > `risky-local-variable' mechanism.  My idea is that only defcustom'ed
>     > variables could be specified in a local variables list, and that the
>     > :type together with the specified value would determine whether a
>     > variable is safe.
>
>     This won't work for defcustoms that haven't been loaded yet.
>
> It would probably refuse to bind the variable in such a case, but is
> that really a problem in practice?  A file local binding is useful
> because the variable is going to do something, which means the package
> that obeys it needs to be loaded.

It might be a problem for variables associated with minor modes that
the user may only turn on after visiting the file.

Re: Instead of `risky-local-variable'

[Kim F. Storm]

Luc Teirlinck <teirllm@dms.auburn.edu> writes:

> Richard Stallman wrote:
>
>    I'm considering a more reliable replacement for the
>    `risky-local-variable' mechanism.  My idea is that only defcustom'ed
>    variables could be specified in a local variables list, and that the
>    :type together with the specified value would determine whether a
>    variable is safe.
>
>    What do people think?
>
> I personally think that this sounds like a pretty radical change,
> supposedly "relatively shortly" before a release.  Any reason why this
> can not wait till after the release?

I don't even want to think about this before the release!

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: Instead of `risky-local-variable'

[Richard M. Stallman]

    I personally think that this sounds like a pretty radical change,
    supposedly "relatively shortly" before a release.  Any reason why this
    can not wait till after the release?

It is a serious security problem.  We should fix it before the release.