Re: Socket inheritance: A tale of two vendors


Chuck Dillon wrote:
> googlegroups@xxxxxxxxxx wrote:
>
> > So, my questions are:
>
>
> > 2) Does it sound like I've got a clear grasp of the problem, or is
> > there some other possibility I've overlooked?
>
> Are you within the margins in how you use the software?

Not doing anything funny. This is totally normal use.

> If the vendors
> aren't interested in fixing the problem it could mean they don't
> acknowledge that there is a problem.

Certainly. That doesn't mean they're being sensible. My experience
has been that they're mostly interested in closing helpdesk tickets,
not fixing software flaws.

Initially, they did agree that there was a problem, but tried to blame
the OS for the inheritance problem. Now that it's clear the origin is
the code, they're not so sure it's a problem anymore :-)

> That they consider you outside of
> their operational expectations or otherwise abusing their software. If
> you are in the margins then the fault lies with you and you aren't
> asking for a fix you are asking for a favor.

This is not the case.

The problem probably hasn't been seen before because:

1) Both packages are expensive, rarely used commercial products, which
are even more rarely used together.
2) Most users of these products are change averse. They do not tweak
parts of their environment frequently, and probably restart
*everything* when they do so.
3) Some other client applications of "daemon" correctly set
close-on-exec, making "daemon"'s sloppy socket handling irrelevant.

> > 1) Does the blame lie equally with both vendors, or is one of them
> > slightly more wrong?
>
> Assuming you are between the margins. If B's API and docs provide a
> mechanism to handle your scenario and A failed to utilitze it the fault
> is with A. Otherwise, B assumes responsibility for the functionality
> provided by their API and the fault is with them.

Thank you, Chuck.

My reading of that comment is: If B advises setting of close-on-exec,
then A should have done so -- otherwise, I will put pressure on B for
their failure to close inherited sockets as prescribed by Stevens.

/chris marget

.

Relevant Pages