summaryrefslogtreecommitdiff
path: root/lisp/gnus/nnheader.el
diff options
context:
space:
mode:
Diffstat (limited to 'lisp/gnus/nnheader.el')
-rw-r--r--lisp/gnus/nnheader.el10
1 files changed, 9 insertions, 1 deletions
diff --git a/lisp/gnus/nnheader.el b/lisp/gnus/nnheader.el
index 1b6ec636734..7df5ecae205 100644
--- a/lisp/gnus/nnheader.el
+++ b/lisp/gnus/nnheader.el
@@ -74,7 +74,15 @@ Integer values will in effect be rounded up to the nearest multiple of
(defvar nnheader-read-timeout
(if (string-match "windows-nt\\|os/2\\|emx\\|cygwin"
(symbol-name system-type))
- 1.0 ; why?
+ ;; http://thread.gmane.org/v9655t3pjo.fsf@marauder.physik.uni-ulm.de
+ ;;
+ ;; IIRC, values lower than 1.0 didn't/don't work on Windows/DOS.
+ ;;
+ ;; There should probably be a runtime test to determine the timing
+ ;; resolution, or a primitive to report it. I don't know off-hand
+ ;; what's possible. Perhaps better, maybe the Windows/DOS primitive
+ ;; could round up non-zero timeouts to a minimum of 1.0?
+ 1.0
0.1)
"How long nntp should wait between checking for the end of output.
Shorter values mean quicker response, but are more CPU intensive.")