--- Begin Message ---
Subject: |
Re: Patch for NIS+ |
Date: |
Wed, 12 Dec 2001 13:39:01 -0500 (EST) |
Realizing that 10 retries with 1 sec increment is arbitrary and
somewhat aggressive, I've made them overridable macro definitions.
Ideally, these should be settable at program start-time or at least
boot-time, not compile time, but I don't know the right place to put
such parameters. (Solaris has two retries with 1 sec timeout increment
both hardcoded).
Setting __NIS_PING_RETRY to 3 (or even 2) sounds reasonable. I've
never had witnessed two failures in a row. Increasing timout step
does not seem to be a good idea in case of a busy network which drops
packets.
Please let me know if I can be of futher assistance.
--
Alexander Belopolsky | Renaissance Technologies Corp.
600 Route 25A | East Setauket | NY 11733-1249
PH: 631-444-7125 | FAX: 631-444-7009
Ulrich Drepper writes:
>
> This patch is severely punishing those which don't have a crappy
> network setup when they really have problems at some point. Instead
> of the 5sec timeout they now face a 95sec timeout, for every single
> programs which uses the code. This is not acceptable.
>
> The code itself, after some moving and cleaning up, seems usable but
> I'd rather see not so many runs of the loop (at most three I guess)
> with larger steps in the timeout value. maybe 5, 7, and 10secs, or 4,
> 8, 12sec.
>
> Ideally this should be a configuration parameter so that only those
> who inflict the problem on themselves are punished.
>
> --
> ---------------. ,-. 1325 Chesapeake Terrace
> Ulrich Drepper \ ,-------------------' \ Sunnyvale, CA 94089 USA
> Red Hat `--' drepper at redhat.com `------------------------
>
--- End Message ---