We'd rather not keep DNS lookups on. But our users demand it. 1.3 will
ship with the lookups turned off by default. In any event, if all you
want is log files with names in them then 1.3 should have a better
facility for that -- we will make the piped logger children reliable
(restartable), right now I wouldn't touch 'em at all. So you can do the
looking outside of the request-path.
Another reason people want DNS along with the hit is to serve ads based on
"domain splits" -- i.e. a different ad to *.edu and *.com. Don't argue
the logic of this, you haven't seen the stupidity of what ad sales folks
will argue for.
Dean
On 8 Jul 1997, John Henders wrote:
> In <Pine.LNX.3.95dg3.970707201058.23100T-100000@twinlark.arctic.org> dgaudet-list-linux-kernel@arctic.org (Dean Gaudet) writes:
>
> >.... Threaded apache on unix
> >won't happen for a while, we're hoping to learn from the NT experience.
> >There are other arguments for delaying threaded unix code... in particular
> >the severe lack of third party libraries that support threading (such
> >as dns, various database libraries, perl).
>
> DNS can be a real killer as Netscape has shown. If you are trying to
> keep reverse lookups on, piping all the requests down a single channel
> backlogs everything majorly. Would spinning off a seperate thread to do
> the lookup work?
>
> When you're doing web hosting it's difficult to turn off DNS lookups as
> clients want that info in their reports, and most report generator slow
> down by a huge amount if you enable lookups. (getstats went from under a
> second to 52 minutes on one test I did) The only other solution I've
> seen is a hack that preprocesses the log and does the lookups, caching
> any results it finds.
>
> --
> Artificial Intelligence stands no chance against Natural Stupidity.
> GAT d- -p+(--) c++++ l++ u++ t- m--- W--- !v
> b+++ e* s-/+ n-(?) h++ f+g+ w+++ y*
>
>