From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: Date: Thu, 7 Sep 2006 19:54:49 -0400 From: "Russ Cox" To: "Fans of the OS Plan 9 from Bell Labs" <9fans@cse.psu.edu> Subject: Re: [9fans] i need someone to pinpoint a sample program for me : ) In-Reply-To: <08e35d54e7ce58abf3f30a8edd7d2ffb@vitanuova.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <08e35d54e7ce58abf3f30a8edd7d2ffb@vitanuova.com> Topicbox-Message-UUID: b097f010-ead1-11e9-9d60-3106f5b1d025 > i have thought in the past that the man pages could make it > clearer how ref counts interact with clunk - i.e. that a client will never > send a clunk while there are outstanding transactions on the fid. > this (i think) falls naturally from the way things are implemented > in plan 9, but might not be obvious to those writing servers (or > new clients). i don't actually know how ref counts interact with clunk. certainly the plan 9 kernels don't send clunks of fids with transactions outstanding, but there is nothing in section 5 that says this is a requirement, and even if there were, robust servers would do well to detect that case anyway. much of the ref counting in lib9p exists for exactly this purpose, to do something sensible even when the client doesn't. russ