Re: [PATCH] sctp: if backlog is 0, listening shall not be deactivated.

[Vlad Yasevich]

Alan Cox wrote:
>> SCTP API simply chooses to ignore the "may".  It is still fully compliant
>> with POSIX in this regard.
> 
> Linux chooses the interpretation that zero means one connection. Having a
> single protocol variant do different things is not nice because  a lot
> of code is designed to handle multiple address families and expect the
> same non address handling behaviour.
> 
> So while it may be able to claim posix compliance, its not Linux
> behaviour, and its relying on a specific interpretation of posix not
> being used by the OS.
> 
> At the very least the current behaviour of the SCTP code is plain rude.
> 

I will submit a requirest to change this behavior in the spec, but I am not
terribly optimistic.  This has been specified for a very long time and there
might be applications taking advantage of the ability to shut listing off.

At this time, let's leave this as is.  A well written application should specify
the listen backlog anyway, otherwise it's depending on the "may" language in the
Posix spec and will not get consistent behavior across different systems.

-vlad
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/