home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!enterpoop.mit.edu!senator-bedfellow.mit.edu!athena.mit.edu!jik
- From: jik@athena.mit.edu (Jonathan I. Kamens)
- Newsgroups: comp.unix.admin
- Subject: Re: mail servers and sendmail -- disk space vs. speed vs. file table trade-offs
- Date: 7 Jan 1993 15:54:01 GMT
- Organization: Massachusetts Institute of Technology
- Lines: 35
- Distribution: world
- Message-ID: <1ihjmpINN38i@senator-bedfellow.MIT.EDU>
- References: <1iesvgINN4l0@senator-bedfellow.MIT.EDU> <3584@aegis.or.jp>
- NNTP-Posting-Host: pit-manager.mit.edu
-
- In article <3584@aegis.or.jp>, davidg@aegis.or.jp (Dave McLane) writes:
- |> Ralph Sims
- |> <ralphs@halycyon.com> sent me a trick version of mini-inews that
- |> didn't just send across the network to the real inews on the server,
- |> but put the article to be posted in a 'printer queue' and then
- |> returned to the user. The code is quite small (8K) and I could post
- |> it here if you/anybody was interested.
- |>
- |> Perhaps you could use something similar to queue the requests to
- |> sendmail.
-
- When you've got users posting small numbers of messages from multiple client
- hosts, this is fine.
-
- However, all of the messages in this case are outbound messages from a single
- host. Sendmail already *has* the functionality you suggest, i.e., the queuing
- delivery mode that I mentioned in my previous posting. As I said before, I
- can't really use that because if I do, a large number of requests in a short
- period of time can fill the disks on my mail server.
-
- I guess that if I wanted to *completely* parallel the news posting model you
- suggest in my server, then I would have to set it up to refuse delivery of
- incoming requests, at the sendmail level, when there aren't currently enough
- resources to process the requests. That's not really possible, because it's
- not *possible* to know when there are going to be enough resources available to
- process a request. In the case of the request that caused me to notice this
- problem in the first place, it was a single request that overwhelmed the
- workstation, and I couldn't have known that the request would be
- resource-consuming without parsing it and trying to deliver (since only then
- would I have discovered the amount of data being transferred and possibly the
- fact that the sender's SMTP server was very slow.
-
- --
- Jonathan Kamens jik@MIT.Edu
- Aktis, Inc. Moderator, news.answers
-