You gave somewhat more detail then I can parse this late.
Simple point, we have run into the 5 pending connection limit on NT
workstation 4.0 in the Java world. Its real, it exists. Where it is
implemented exactly in the OS I have no clue.
JK
At 05:53 PM 3/30/01 -0500, Jon A. Lambert wrote:
>Jeff Kesselman wrote:
> >
> > Keep in mind the fact that MS purposefully crippled the networking in
> > everything but NT server. NT Workstation has a limit of 5 pending
> > connections to a socket. I don't know about the 95 codebase but I suspect
> > its similar.
> >
> > This means if more then 6 people try to get connected to your MUD at once,
> > everyone after the sixth will get rejected as if there was no socket open
> > on the MUD machine.
> >
>
>The Winsock2 stack is identical on all versions of Windows that support
>it (i.e Win95/98/ME/NT/2K client or server versions). There is a 32K
>limit on the numbers of simultaneous TCP/IP connections on ALL versions
>of the Windows OS that support Winsock2.
>
>There are differences in the tradition *nix TCP implementation and the
>BSD/Winsock2 TCP implementation on what is sent to the client when the
>"pending queue of connections" is full. Anyhow the Winsock2
>implementation attempts to mimic BSD not *nix. The "pending queue of
>connections" is _not_ to be confused with a "connection limit", as all
>TCP stacks implement limits on the size of this queue. Obviously since
>Cold Dark is/was(?) running on FreeBSD, any problems with how BSD handles
>listen/accept when the queue is full would have already cropped up.
>And since they haven't to my knowledge, this is a red herring.
>
>You may also confusing seat license limits on the server software that
>ships as part of NT. The server software that ships with client OS
>versions is different than that shipped with server OS versions. Those
>connection limits are imposed by the server software and _not_ the TCP
>stack. ColdX/Genesis is obviously not subject to any crippled connection
>limitations. This is easy enough to test using any Windows version
>that can run Winsock2.
>
>--
>--* Jon A. Lambert - TychoMUD Email:jlsysinc@ix.netcom.com *--
>--* Mud Server Developer's Page <http://tychomud.home.netcom.com> *--
>--* If I had known it was harmless, I would have killed it myself.*--
Jeff Kesselman
Staff Engineer -- SUN SSG Lab
Co-author of Java(tm) Platform Performance
ISBN: 0-201-70969-4
|