We encountered a similar problem on NT. Charlie Caro post a message to list
a week ago confirming it was the same problem. NT seems to accept the
connections above 256, ie the client apps do not get an error, but neither
do the get any response from the server.

Regards Andre

-----Original Message-----
From: Jason Chapman <jason@...>
To: IB-Architect@egroups.com <IB-Architect@egroups.com>
Date: Wednesday, July 05, 2000 10:43 PM
Subject: Re: [IB-Architect] Interbase connection limit and Support
relatedProblems


What about NT?

My largest customer has just blown 256!

JAC.

----- Original Message -----
From: Jim Starkey <jas@...>
To: <IB-Architect@egroups.com>; <IB-Architect@egroups.com>
Sent: 05/07/2000 4:54 PM
Subject: Re: [IB-Architect] Interbase connection limit and Support related
Problems


> At 04:53 PM 7/5/00 +0200, Andre Mostert wrote:
> >
> >Interbase should be able to support more than 256 connections. The most
> >disconcerting fact is that in the last two weeks we
> >have emailed Ann Harrison and copied Paul Beach twice just to ascertain
what
> >progress has been made or when we can expect
> >resolution to our problem. We have not enjoyed the curteousy of a reply.
We
> >realize that you are busy with ver 6 and probably
> >with preparations for the conference. We however are sitting with 18
centres
> >on our new app, and the balance awaiting
> >resolution from Interbase and our software house.
> >
>
> I'll let Ann handle the business issues and take on the architectural
> ones.
>
> You're not going to like the answer, but the problem is Sun's. InterBase
> can work around the Sun's problem, but only Sun has the ability to
> fix it.
>
> Here is the problem (from /usr/include/stdio.h):
>
> /* @(#)stdio.h 1.16 89/12/29 SMI; from UCB 1.4 06/30/83 */
>
> # ifndef FILE
> #define BUFSIZ 1024
> #define _SBFSIZ 8
> extern struct _iobuf {
> int _cnt;
> unsigned char *_ptr;
> unsigned char *_base;
> int _bufsiz;
> short _flag;
> char _file; /* should be short */
> } _iob[];
>
> The field _file holds the FD (file descriptor). On Solaris (like
virtually
> all Unix systems), sockets and file descriptors come out of the same
> number space, and are assigned sequentially, starting with 0. A
> "char" field can only hold a number up to 255 (127 if you want to
> be picky). If you start with stdin, stdout, and stderr tying down
> 0, 1, and 2, add a database file, that leaves a max of 252 connections
> before the next file open overflows _file and the world ends.
>
> The obvious workaround is to use a less-stupid set of buffered
> IO functions, which the InterBase guys tried. A little painful, but
> given a choice of open source file handling subsystems, completely
> feasible. This was an imperfect solution as some other limit
> (to the best of my knowledge, undiagnosed).
>
> The bottom line is that exceeded in the inherant limitations of the
> underlying operating systems is a tricky problem. Since anybody
> using the standard Sun libraries is stuck with the 255 file/socket
> limit, who knows what else is going to break.
>
> Since this is an architectural and not a support list, I will just
> point out that HPUX uses two bytes for the file descriptor and
> Linux a full int.
>
>
>
> Jim Starkey
>
> ------------------------------------------------------------------------
> Free, Unlimited Calls Anywhere!
> Visit Firetalk.com - click below.
> http://click.egroups.com/1/5479/6/_/830676/_/962812634/
> ------------------------------------------------------------------------
>
> To unsubscribe from this group, send an email to:
> IB-Architect-unsubscribe@onelist.com
>
>
>


------------------------------------------------------------------------
Where do sports heroes like Derek Jeter, Mia Hamm,
Vince Carter and Peyton Manning hang out? Where else?
Click now and find ‘em all here!
http://click.egroups.com/1/6211/6/_/830676/_/962829799/
------------------------------------------------------------------------

To unsubscribe from this group, send an email to:
IB-Architect-unsubscribe@onelist.com