Re: [Lynx-dev] how can I trace down "pauses"

[Thomas Dickey]

On Tue, 1 May 2007, Virden, Larry W. wrote:

> Hi!
>
> Recently a favorite web site (http://wiki.tcl.tk/) changed http servers.
> Since that occurance, I've been noticing a peculiar behavior with lynx
> (2.8.7dev.4 built on sparc solaris is my primary version).
>
> What I see happen is this - in the first second or so, the page appears.
> In the status message, I see Read 12 of 13 KB of data, 6 KB/sec. or
> perhaps 18 of 22 KB. Then lynx sits for 30 seconds, and eventually
> either gets the data or gives up - I can't tell which.
>
> The weird thing is that if I use 'z' to tell lynx to stop , I look
> around and see what appears to be the entire page.
>
> So, what approach would you take to try to figure out what is going
> on between lynx and the server that is causing the 30 second 'delay'?

hmm - see

        http://wiki.tcl.tk/15781

Looks like the problem is that wub isn't closing its connections in a 
timely fashion (this affects w3m as well).  Perhaps the 30-seconds comes 
from a timer in wub that closes inactive sessions - I don't see anything 
on lynx's side that's forcing a timeout, and 30-seconds seems too low for 
network timeout.

Reading lynx's code, it appears that it relies on the server to close the 
connection after sending data.  I suppose lynx "could" close the 
connection (would be surprised if that's not addressed in an RFC), but 
it's not doing the required bookkeeping.  I think that getting wub to 
behave like other servers is the best route.

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net