Dave McCaldon wrote:
> > Or are you suggesting that all authentication failures (as
> > opposed to other errors such as sending failures) should share an
> > error code regardless of the authentication method used?
> 
> As I understood it, this was Peter's suggestion.

It was meant as an idea, sort of a call for comments. I certainly did
not intend to suggest that it was the prefered solution. Sorry for
the confusion!

The reason that I didn't just commit your patch has nothing to do
with form and everything to do with my gut feeling that this handling
should probably be thought about and discussed some more. I don't
think that sending new variations of the patch is neccessarily the
fastest way forward until we've nailed how it should work. :)


> In any case, if I can get a final opinion from the group on this,
> I'm happy to implement it.

Let's figure it out first. Please be patient and comment a lot on all
the good and bad ideas.


The thought that occured to me before was that we might need to take
a bit of a stand in libssh2 here on whether it should provide error
codes that are close to the protocol errors (which I would prefer) or
whether it should try to be smart and cook protocol error codes into
something that could be expected that application writers would
prefer, or just where we think that protocol errors suck.

Personally I rather like the protocol but on the other hand maybe
it's too low level sometimes?

I absolutely agree that this failure path needs to be covered, but
that doesn't mean it's simple to produce a "final opinion".


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Reply via email to