[IRCServices Coding] Cygnus

Trevor Talbot quension at softhome.net
Thu Apr 11 18:57:29 PDT 2002


On Thursday, April 11, 2002, at 03:06  PM, Kelmar K. Firesun wrote:

>> <griever at t2n.org>:
>>>>      I still haven't heard of a _real_ network with 100k 
>>>> users, but even
>>>
>>> You've never heard of dalnet?
>>
>>      I wasn't aware they had 100k simultaneous users, but then 
>> I haven't
>> gone there in a while.
>
> Current local users: 10288 Max: 25358
> Current global users: 120185 Max: 135705
>
> I don't think you'd want to go back to Dalnet anytime soon:
> -twisted.ma.us.dal.net- *** Autokilled for [exp/ident] Enable 
> ident in your
> client. Send email to exploits at dal.net with a subject of 
> [exp/ident] for
> more
> details. [AKILL ID:969137152K-a] (2002/04/11 14.58)
>
> *Tries not to go on tirade about how using IDENT for security is a Bad
> Idea(tm)*

Don't worry; it's not being used for security.

>>>>      Also, it's worth noting that Services' I/O engine has 
>>>> been completely
>>>> redesigned for 5.0, so any comparisons done with 4.x don't count.
>>>>
>>> Actually, if I remember correctly 5.0 was slower. May be 
>>> because of added
>>> features
>>
>>      As mentioned above, commands have more overhead (mostly 
>> because they
>> use callbacks), so any comparisons done with 4.x don't count 
>> even more.
>
> That and loadable modules, which would have the additional overhead of
> context switching.  *shudder*
>
>>> It's of point to note that both versions of ircservices sent a 
>>> lot more
>>> to the server. Any idea what this could be?
>>
>>      Because they're friendlier and want to chat with the server more?

While we're on this topic, I took a look at Cygnus' 0.0.2 socket code.
It appears to use a blocking socket, and I see possible issues with
receive buffering.  Did it handle everything it was supposed to?

More details really are needed for decent comparisons.

-- Quension