Re: two imap.el patches

[Simon Josefsson <jas@extundo.com>]

On 20 Jun 2001, NAGY Andras wrote:

> On Tue, 19 Jun 2001, at 21:10, Simon Josefsson <jas@extundo.com> wrote:
>
> > I mean, if the default is nil, it will not work at all if I have the
> > old imtest, and will work but with line length problems (that's all
>                                     ^out, you mean?

Right. Oops.

> > right?) with the new imtest.  If the default is t, it will work but
> > have line length problems regardless of imtest version.
>
> Hm, I not sure about these line length problems... the major reason I
> was looking for a solution not using a pty was the insecure pty
> allocation method on some systems (notably AIX 4.3, which I am using
> :).  I don't know if using a pty might introduce line length problems.

Aha.  Using a pty makes us hurt from OS buffers on line length limits,
and some OSs (Irix come to mind) have 256 character line length buffers.
So using a pipe probably (but I'm not sure) means we don't have that
problem anymore, and that's a good thing.  If it improves security as
well, great.

> imtest version / p-c-t       nil                              t
> <2.0.14                      does not work not at all         insecure
> 2.0.14+                      fine and secure                  insecure

Right.  So it's a tradeoff between providing almost working solutions for
as many people as possible, or to provide a completely working and secure
solution for fewer. But then, people using GSSAPI or Krb4 is probably not
completely clueless and should be able to figure these things out, so nil
is a good default here. I added some information to gnus.texi about this.