Eggdrop1.6.6: Timeout/EOF ident connection

[egghead]

In testing a TCL script, the following was observed.

First eggdrop1.6.6 is started (SunOS, TCL 8.0) with 3 modules:

Eggdrop v1.6.6 (C)1997 Robey Pointer (C)2001 Eggheads
[18:36] --- Loading eggdrop v1.6.6 (Mon Jan 7 2002)
[18:36] Listening at telnet port 32000 (users)
[18:36] Module loaded: notes (with lang support)
[18:36] Module loaded: blowfish
[18:36] Module loaded: dns
[18:36] Userfile loaded, unpacking...
[18:36] === Eggchat: 0 channels, 1 user.
Launched into the background (pid: 10865)

Note: this bot does not connect to IRC or other bots. It runs limbo for testing purposes. ident-timeout is set to 10.

The following piece of TCL is loaded:

Code: Select all

# listen on port 4000 

listen 4000 script observescript 
   proc observescript { idx } { 
   control $idx observeproc
   return 0 
} 

proc observeproc { idx text } { 
   putlog "OBSERVE: ($idx) $text" 
   return 0 
} 

When a connection comes in from a Netscape client, the following is reported on the bot:

[18:46] Telnet connection: netscapehost/2711
[18:46] Timeout/EOF ident connection

Note: these 2 lines follow each other immediatly without any delay.
My first question is: what does this "ident connection" do? Is it really needed? Second question: can it be disabled, or are other eggdrop settings needed?

Then a connection is made using Lynx, which is to submit some text using a HTML form and POST.

What then happens is:

[18:48] Telnet connection: lynxhost/4568
(then a delay of about 10 seconds ( ident-timeout?))
[18:48] Timeout/EOF ident connection

After this timeout, all lynx client header information is received correctly. Now one of the interesting points is that if Lynx then kills the connection (by terminating Lynx itself ), this termination is not processed by eggdrop. No empty string is sent to the observeproc. I.e. eggdrop still thinks the connection exists.

The 3rd issue is the following: a connection with Lynx is made, and right thereafter a connection is made by Netscape. In this case the following is observed:

[19:04] Telnet connection: lynxhost/4995
[19:04] Telnet connection: netscapehost/2718
[19:04] Timeout/EOF ident connection
[19:04] Timeout/EOF ident connection

In this case no delays are visible. Upon the Netscape connection, both "Timeout/EOF" appear followed by the info submitted by Lynx and Netscape respectively.

I won't describe here that by making/disconnecting two or more times repeatedly after one another with Lynx, things got "blocked". i.e. all information submitted by Lynx was not sent to the observeproc procedure. Only after making one Netscape connection, all this data was sent to the observeproc.

All in all, this "Timeout/EOF ident connection" seems a bottleneck. Any suggestions to circumvent it are greatly appreciated.

<font size=-1>[ This Message was edited by: egghead on 2002-01-07 19:24 ]</font>

[Petersen]

reduce your $ident-timeout to a low value. Anything routed through eggdrop's own listen sockets will always try to get a valid identd so it can check the user@ of the connecting client for auth purposes.

[ppslim]

There was a change made to eggdrop some time ago, to facilitate this, yet it failed to do the job, and was then removed (as there is no trace of it in the code now).

However, what you have observed is a bug, which I have allready reported, but have seen nobody report back with anything sugesting a fix.

This "Timeout/EOF ident connection" happens with every incoming connection, when the bot is running in Limbo, only when the bot is connected to IRC, whill this go away. As you can see, this should not happen.

Lowering the timeout value to 1 didn't help me much.

My sugestion would be to report these observations to eggheads@eggheads.org, and also ask for some sort of method to disable identd request's for listening sockets, with some sort of flag.

That, or can stdragon or guppy, shed some light in to either of these issues.

[egghead]

Ok, thanks. I was about to report how I started out by setting the ident-timeout to 0. In the hopes that eggdrop then would understand to disable the ident connection. Setting it to this value doesn't solve too much though: the earlier mentioned timeout of 10 seconds reduces to a lower value, but the Lynx connections still get "blocked".
But after reading ppslim's comments I'll refrain from mentioning all this

So, is it correct that in higher versions of eggdrop this ident has been removed?
Or is this another argument to use TCL sockets?

Don't want to start an off-topic subject, but as a sidenote, I dont see much added value of "checking/authenticating" the user@. This will only work for checking connections from correctly (and honestly managed boxes.

<font size=-1>[ This Message was edited by: egghead on 2002-01-07 20:17 ]</font>

[ppslim]

At this moment in time, yes it is antoher argument to move to Tcl sockets.

This has not bee fixed in new version of eggdrop, and is doubtful to be fixed until 1.8.