httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From (Robert S. Thau)
Subject Re: lights, cameras, action
Date Thu, 09 Mar 1995 13:25:39 GMT
   Date: Thu, 9 Mar 95 17:43 GMT
   From: (David Robinson)

   B1c) I have another patch (to translate_name) which fixes another hole in a
   similar manner to the NCSA patch.

Now on my list.

   > B2) Don't set SO_LINGER on client sockets.
   Just out of interest, why did httpd do this?

Damned if I know.  NB the CERN server used to do it too by default,
but it's now disabled on most configurations (though you can turn it
on in the rule file).

   >P2) Don't do kernel read()s of one character, when reading MIME
   >    headers from clients, or script CGI headers.
   >    drtr and rst have their own versions; drtr's handles POST, but
   >    looks like it might have somewhat more overhead in the simple GET
   >    case.

   My patch isn't necessarly slower than rst's for GETs; it does 3p
   system calls where p = no of data packets (select,recv,read per
   packet), whereas rst's does 4 + 3p calls (4 1-byte reads +
   signal,alarm,read per packet).

That's probably good to first order; I've edited the note. 



View raw message