TCP Wrappers

[Charles Wilson]

Corinna Vinschen wrote:

> It clarifies.  I was just confused how inetd comes into play here.
> 
> Just consider that if the usage of a tcp wrappers dll actually
> requires us to change sshd code, I'd better live with static linking.
> libwrap.a isn't that big.


All true.  However, if the API of libwrap needs to change in order to 
support building shared libs on certain platforms (cygwin, native 
windows), then that'll happen at the upstream tcp maintainer level.

Which, IF he did that, would cause a ripple effect as the various 
changes to client code were added all over the opensourcesphere. 
Fortunately, given the .so versioning tools, this wouldn't break 
existing builds of, say, openssh on linux.

But to get this patch to the point where the upstream maintainer would 
even consider it, someone "down here" -- like Prentis -- needs to 
develop and test the patch.  (That's what I did w.r.t. ncurses, 
readline, etc).  Then, push it upstream and see if it'll be accepted.

Whether cygwin's version of openssh adapts to the dllized API or not is 
of course up to Corinna.  (Although, this does present one problem: if 
Corinna doesn't want to use the modified API, then Prentis will have to 
maintain two APIs -- one for the DLL and one for the statlib, using 
#ifndef WITH_LIBWRAP_DLL guards in tcpd.h....blech)

--Chuck