2.9p2 behaves different from 2.5.2p2 on tunneling issue

[Markus Friedl]

On Sat, Sep 15, 2001 at 01:57:02AM +0200, Corinna Vinschen wrote:
> On Fri, Sep 14, 2001 at 07:54:05PM +0200, Markus Friedl wrote:
> > On Thu, Sep 13, 2001 at 02:07:02PM +0200, Corinna Vinschen wrote:
> > > On Thu, Sep 13, 2001 at 11:35:42AM +0200, Markus Friedl wrote:
> > > > On Wed, Sep 12, 2001 at 02:33:59PM +0200, Corinna Vinschen wrote:
> > > > > On Wed, Sep 12, 2001 at 02:00:53PM +0200, Markus Friedl wrote:
> > > > > > is the server running 2.9, too?
> > > > > 
> > > > > No, just in case you missed my previous mail, it's running
> > > > > F-Secure 2.0.12 on some Sparc running SunOS.
> > > > 
> > > > ok, i think that ssh.com < 2.1.0 behaves different from
> > > > newer releases.
> > > > 
> > > > newer ssh.com servers keep the connection, older servers
> > > > tear the connection down.
> > > 
> > > But, uhm, why does that result in different behaviour of 2.9p2
> > 
> > because 2.5 had a 'broken' implementation like the older ssh
> > servers. it was 'fixed' in 2.9, now you see this behaviour
> > against older, broken servers. the problem is that
> > the drafts for the protocol have changed since 1997.
> 
> Shouldn't 2.9 behave `broken' when it knows that it communicates
> with an old `broken' server?
> 
> > > in contrast to 2.5.2p2 as client using the same server?  And
> > > how can I workaround that using 2.9p2?  I'm not that keen to
> > > stick with 2.5.2p2 on that machine...
> 
> Otherwise there's neither a workaround nor a sort of a compatibility
> mode in 2.9.  Which means I'm stuck with 2.5.2 unless the server
> side decides to upgrade :-(

look at compat.c, we have many emulations of 'broken' code.
i'll look into this, but i think it's too ugly.