Re: [Quickfix-developers] FAST FIX

[Shawn Y. <sya...@ge...>]

Basically, we disable sequence numbers.  FIX on CME runs over TCP/IP, =
which is supposed to be a reliable protocol.  So we aren't too worried =
about dropped messages.

All that the sequence number feature did for us was prevent us from =
reconnecting to CME after a problem.  There are any number of good or =
bad reasons that your sequence numbers might be off, including:

1) You are developing new trading software, and had to delete your =
QuickFIX state files to reset QuickFIX for some reason.
2) You moved your trading software to a new machine, but didn't think of =
copying over the QuickFIX state files.
3) Your trading software crashed, which is hopefully rare in mature =
software, but will happen, and which is common in new software, when you =
need your connection logic to be at it's easiest.
4) Somebody else in your company mistakenly logged into your session =
from some other machine.  If they do this while you are logged in, CME =
silently disconnects you, which is another whole story about painful =
troubleshooting.
5) CME is off in their sequence numbers for unknown reasons.  Maybe they =
rebooted their simulation environment without asking your permission.  =
Exchanges don't tend to care about your convienence.

After a crash, your traders have already phoned the exchange and =
manually maintained their orders as needed.  So you aren't concerned =
with the perfect integrity of the data flow.  Now your traders, who are =
angry with you over the crash, just want you to bring the software back =
online so they can get back in the market.  But wait, FIX+QuickFIX+CME =
are here to help you.  When you restart your trading software, it and =
CME will refuse to talk to each other because they helpfully noticed =
that the sequence numbers are off.

But you don't care!  You just want your traders to get back online.

So we disable sequence numbers.  How many hours I wasted trying to =
figure out why CME and my code were speaking to each other...

Sequence numbers left a bad taste in my mouth for the FIX protocol.  I =
realize part of the problem may have been poor documentation, and part =
of the problem may have been CME's quirky FIX implementation, but it's =
the truth.  Connecting sophisticated processes to each other can be hard =
enough without the protocols actively trying to prevent you from doing =
it.

Another problem were the session start and end times.  Most networking =
protocols don't require you to predict the future before you are allowed =
to connect.  How long will you be connected?  We now set a start time of =
one second after midnight on Sunday morning, and a stop time of one =
second before midnight Saturday night.  Before I figured those settings =
out, we couldn't connect, and it wasn't clear why not.

It would be nice if QuickFIX could help smooth over these FIX and CME =
speedbumps.


Shawn Yarbrough
Software Architect
Gelber Group, LLC
(312) 692-2819
sya...@ge...


THE MESSAGES AND DOCUMENTS TRANSMITTED WITH THIS NOTICE CONTAIN =
CONFIDENTIAL INFORMATION BELONGING TO THE SENDER. IF YOU ARE NOT THE =
INTENDED RECIPIENT OF THIS INFORMATION, YOU ARE HEREBY NOTIFIED THAT ANY =
DISCLOSURE, COPYING, DISTRIBUTION OR USE OF THE INFORMATION IS STRICTLY =
PROHIBITED. IF YOU HAVE RECEIVED THIS TRANSMISSION IN ERROR, PLEASE =
NOTIFY THE SENDER IMMEDIATELY.



-----Original Message-----
From: Oren Miller [mailto:or...@qu...]
Sent: Monday, July 23, 2007 4:00 PM
To: Shawn Yarbrough
Cc: Brian Erst; Caleb Epstein; qui...@li...
Subject: Re: [Quickfix-developers] FAST FIX


Do you have information about what you are doing with sequence number =20
handling?

--oren

On Jul 23, 2007, at 3:52 PM, Shawn Yarbrough wrote:

> We've been running QuickFIX on CME for about a year and a half.  I =20
> had to disable QuickFIX's message validation dictionary, and =20
> override some of the sequence number handling, but otherwise it =20
> works OK.
>
>
> Shawn Yarbrough
> Software Architect
> Gelber Group, LLC
> (312) 692-2819
> sya...@ge...
>
>
> -----Original Message-----
> From: qui...@li...
> [mailto:qui...@li...]On Behalf Of
> Brian Erst
> Sent: Monday, July 23, 2007 3:42 PM
> To: Oren Miller; Caleb Epstein
> Cc: qui...@li...
> Subject: Re: [Quickfix-developers] FAST FIX
>
>
> QuickFIX Documentation: http://www.quickfixengine.org/quickfix/doc/=20
> html/index.html
> QuickFIX Support: http://www.quickfixengine.org/services.html
>
>