pcp
[Top] [All Lists]

Re: pcp updates: pmwebd security, pmdaroot, libpcp, qa

To: Nathan Scott <nathans@xxxxxxxxxx>
Subject: Re: pcp updates: pmwebd security, pmdaroot, libpcp, qa
From: fche@xxxxxxxxxx (Frank Ch. Eigler)
Date: Wed, 11 Nov 2015 07:26:14 -0500
Cc: PCP <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <689965523.9103364.1447191909906.JavaMail.zimbra@xxxxxxxxxx> (Nathan Scott's message of "Tue, 10 Nov 2015 16:45:09 -0500 (EST)")
References: <190227633.9098813.1447191336585.JavaMail.zimbra@xxxxxxxxxx> <689965523.9103364.1447191909906.JavaMail.zimbra@xxxxxxxxxx>
User-agent: Gnus/5.1008 (Gnus v5.10.8) Emacs/21.4 (gnu/linux)
Hi -

> [...]
>       libpcp: allow exclusive socket flag to be passed via hostspec too

Can we discuss this PM_CTXFLAG_EXCLUSIVE idea?  How is a pmapi user
supposed to know whether it should be used?  They have no idea how
libpcp decides to share sockets to pmcd.  They have no idea what the
implications of sharing are, in terms of which pmapi operations are
affected intuitively or counterintuitively.  Even if they did, socket
sharing is supposed to be a transparent optimization, beneath the api
level.

If socket sharing is not working right, let's disable it or fix it,
not expose its mysteries to the user.  It is an api design mistake to
make users cargo-cult a workaround around problems even we don't seem
to understand.


- FChE

<Prev in Thread] Current Thread [Next in Thread>