Re: [kitten] GSS-API and timeouts

[Nico Williams <[e-mail]>]

On Wed, Apr 4, 2012 at 12:20 PM, Simon Josefsson <[e-mail]> wrote:
> Nico Williams <[e-mail]> writes:
>
>> If you don't want to tackle async init/accept_sec_context now then I
>> suggest that you use threads and thread cancellation.
>
> And not have any timeout in the implementation of the mechanism itself?

Right.  Why not?  You'd have to have cancellation handlers and so on,
but then, proper pthread use requires it anyways.  Not that anyone
bores making libraries cancel-safe -- it's a lot of work, and
pthread_cacnel() is rarely ever used.

On the other hand, both MIT and Heimdal have added the sorts of
interfaces you proposed without any standardization effort, so nothing
should stop you from doing the same.

My problem with global settings is that they are global, but we know
we have apps that load libraries that load libraries that load
libraries that use GSS such that we end up having multiple GSS apps in
one process.  This is why I dislike global settings.  My solution,
when I get around to finishing it, will be a variant of the "pgss"
proposal... (if you'd like to review the design you can, it's in my
'pgss' branch of my github Heimdal fork).

Nico
--