home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.isis
- Path: sparky!uunet!think.com!rpi!batcomputer!cornell!ken
- From: ken@cs.cornell.edu (Ken Birman)
- Subject: Re: spooler usage
- Message-ID: <1993Jan9.140019.23960@cs.cornell.edu>
- Organization: Cornell Univ. CS Dept, Ithaca NY 14853
- References: <QfHZ0bu00hNP9DzF9=@cs.cmu.edu>
- Date: Sat, 9 Jan 1993 14:00:19 GMT
- Lines: 38
-
- In article <QfHZ0bu00hNP9DzF9=@cs.cmu.edu> Sean.Levy@cs.cmu.edu writes:
- >I have a very simple question about the use of the spooler. After having
- >read the spool(3) man page, one thing is not clear to me: what does a
- >reply() call do when given a message that came (either via play-back or
- >play-through) from the spooler?
-
- I'll have a look at the code, but I think the reply would go back to
- the spooler where it would be discarded. The real point is that
- since the sender is not waiting for replies, having called spool
- to send the message rather than bcast, even if it went back to the
- sender the sender would ignore the reply.
-
- So, you'll want to make a table of pending operations. When you send
- to someone over your slow line, also add a new pending op entry to the
- table and include a reference to it in the message.
-
- Implement a reply "manually" by spooling a message back in the other
- direction that includes the op reference explicitly. Pair them up,
- and you have essentially done what collect-replies does now (except
- that your version won't handle nullreply, abortreply, multiple replies
- from a group, or failures :-( ) This is what we meant by a "call back"
- style in the documentation.
-
- The problem is that leaving a task waiting for hours or days just
- doesn't make a lot of sense, and the basic scheme for replies keeps
- a task around. Long delays associated with spooler interactions to
- infrequently run servers or to the WAN code just don't match with
- a synchronous request-reply style of coding. We originally thought
- about doing it in a more transparent way, but ended up with this
- instead.
-
- In Horus, which will focus more on news, we may have another look
- at this whole issue...
-
- --
- Kenneth P. Birman E-mail: ken@cs.cornell.edu
- 4105 Upson Hall, Dept. of Computer Science TEL: 607 255-9199 (office)
- Cornell University Ithaca, NY 14853 (USA) FAX: 607 255-4428
-