Re: [Nmh-workers] Garbage collection

[Anthony J. Bentley]

Hi Ken,

Ken Hornstein writes:
> >The "brokenness" is that OpenBSD simply doesn't implement utmpx, because
> >it's seen as an unsafe and insecure interface. OpenBSD aren't the only ones
> >who feel this way (the musl C library also doesn't support utmpx, for the
> >same reasons: http://www.openwall.com/lists/musl/2012/03/04/4).
> 
> I'd find the "security" arguments more compelling if OpenBSD didn't
> implement utmp, which has (as far I can tell) the same security issues.

Well, I'm just repeating what I've heard elsewhere. I am not familiar
with how utmp{,x} works myself. utmp is no doubt kept around because
existing software demands it, and the implementation is mature enough
to not have known holes.

> >As far as I know the behavior of the utmpx functions are not defined by
> >POSIX either.
> 
> http://pubs.opengroup.org/onlinepubs/009696899/functions/endutxent.html

And on that page utmpx is marked as an XSI extension. To my understanding
those are optional and not required to be implemented in POSIX-compliant
software.

--
Anthony J. Bentley