Subject:
|
Re: "jump.cgi" considered harmful ? (1)
|
Newsgroups:
|
lugnet.admin.general
|
Date:
|
Thu, 23 Nov 2000 15:12:55 GMT
|
Viewed:
|
3923 times
|
| |
|
|
In lugnet.admin.general, Larry Pieniazek writes:
> In lugnet.admin.general, Kevin Loch writes:
> > Are you sure it's not latency in DNS resolution?
> This is almost certainly what a lot of it indeed is. Using jump.cgi
> theoretically doubles the latency since two resolutions are required.
If you're reading via HTTP, then it's only one resolution because your client
will already have resolved www.lugnet.com. If you're reading via NNTP, then
it may be two and it may be one depending on your DNS cache.
> But it's not ALL the delay, some surely, is at the server itself while it
> processes. [...]
> I am curious, however, as to how much of the delay is due to processing on
> the server. [...]
With a typical URL, the script takes .015 seconds (about 1/60 second) of real
time once it receives the HTTP request. Almost all of the overhead is in
launching Perl and compiling its script. (It's not a mod-perl script.)
But Apache pre-forks, and FreeBSD forks quickly anyway, so the total time
shouldn't be more than .016 seconds.
A non-typical URL, such as an Amazon.com URL which contains no associate ID,
may add another 1/10000 second or so. Basically the script just takes the
URL from the QUERY_STRING environment variable, outputs an HTTP Location:
header, then exits. All the logging is done transparently by the normal HTTP
logging mechanisms -- virtually no overhead because it keeps the log
filehandle open all the time.
> Moreover, I would put up with the delay less grumpily if I had
> access to the data, that is, if the delay somehow benefited me (except in
> the meta sense). Finally, I suggest that on the post preview page, it be
> suppressed, since those jumps are not statistically meaningful.
Hmm, good idea.
--Todd
|
|
Message has 1 Reply:
 | | Re: "jump.cgi" considered harmful ? (1)
|
| (...) Did I just say that? Oops. That's wrong. There isn't any jump.cgi in the context of NNTP -- only HTTP. Duh. So it's never more than 1 DNS lookup in the case of NNTP and never more then 2 DNS lookups in the case of HTTP -- and in practice, it (...) (24 years ago, 23-Nov-00, to lugnet.admin.general)
|
Message is in Reply To:
| | Re: "jump.cgi" considered harmful ? (1)
|
| Let me try to be clearer. (...) This is almost certainly what a lot of it indeed is. Using jump.cgi theoretically doubles the latency since two resolutions are required. But it's not ALL the delay, some surely, is at the server itself while it (...) (24 years ago, 23-Nov-00, to lugnet.admin.general)
|
36 Messages in This Thread:
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
