httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chuck Murcko <ch...@telebase.com>
Subject Re: Broken clients
Date Wed, 17 Jul 1996 16:19:03 GMT
Roy T. Fielding liltingly intones:
> 
> > Actually, it's easier to do this in send_http_header():
> > 
> > void send_http_header(request_rec *r)
> > {
> >     conn_rec *c = r->connection;
> >     BUFF *fd = c->client;
> >     const long int zero=0L;
> >     array_header *hdrs_arr;
> >     table_entry *hdrs;
> >     int i, is_agent_mozilla2;
> >     char *agent;
> > 
> >     core_dir_config *dir_conf =
> >       (core_dir_config *)get_module_config(r->per_dir_config, &core_module);
> >     char *default_type = dir_conf->default_type;
> >         
> >     agent = table_get(r->headers_in, "User-Agent");
> >     is_agent_mozilla2 = strncmp(agent, "Mozilla/2", 9);
> > 
> >     ...
> > 
> >     if(!is_agent_mozilla2)
> >         set_keepalive (r);       
> > 
> >     ...
> > 
> >     if (c->keepalive && !is_agent_mozilla2)
> >         bflush(r->connection->client);  /* For bugs in Netscape, perhaps */
> >     
> >     ...
> > }
> 
> Ummm, isn't that redundant?  I mean first not setting keepalive and
> then also checking is_agent_mozilla2 after checking keepalive.  I'd think
> only the first was needed, but I may be confused.
> 
Yes, it is. This isn't the code that would be an official patch.
I'm still testing that here. I get unusual results when I telnet to port
80 and do an HTTP 1.0 request for the root document. As soon as I fix that,
I'll submit a patch for review. I may have stumbled onto the Netscape
3.0b4 (?) core dump problem with this one.

> > This also takes care of John's first point re: keepalive performance.
> 
> Yep, I forgot to mention that -- we can take out the flush entirely
> because is_agent_mozilla2 wouldn't be kept alive [muahahaha...now all
> we need is a picture of a lizard on a roasting stick...].
> 
> BTW, I am inclined to allow any bugs in 3.0beta clients to be visible
> (if they exist), since that is the only way they'll get fixed.  If 3.0 final
> also has the bug(s), then we can kill the whole Mozilla family.
> However, people running commercial sites (i.e., slaves to the wheel)
> may have a different opinion of how best to support their customers,
> so I wouldn't mind having an IFDEF for the string to match and tell
> people to "adjust it accordingly".
> 
Sounds good here. I'll have the graphic ready this weekend. 8^)

chuck
Chuck Murcko	N2K Inc.	Wayne PA	chuck@telebase.com
And now, on a lighter note:
For a man to truly understand rejection, he must first be ignored by a
cat.

Mime
View raw message