[This message was posted by Rolf Andersson of Pantor Engineering <[EMAIL 
PROTECTED]> to the "FAST Protocol" discussion forum at 
http://fixprotocol.org/discuss/46. You can reply to it on-line at 
http://fixprotocol.org/discuss/read/cc3cdf5e - PLEASE DO NOT REPLY BY MAIL.]

All,

the individual extensions are very different in nature.

I believe we should be careful not to include anything in the standard that we 
cannot demonstrate is definitely needed. As Jim commented: "when in doubt leave 
it out".

I also think that we should follow the IETF mantra of "implementation before 
standardization". Preferably at least two implementations that could be 
compared for interoperability.

I'd like to do the following grouping of the proposed features:

1.1.1 scope
- Enumeration data type
- Set data type
- Timestamp data type

1.2 scope
- Bitgroup data type

2.0 scope
- Map operator
- Pmap field placement


The 1.1.1 scope is fairly obvious too me.

It's needed for the FIX over FAST project. None of the 1.1.1 features break any 
existing implementation apart from the added template syntax that can be 
emulated with uInt fields.


The 1.2 scope has limited empirical backing.

Greg Maynard of the ISE has done some analysis and IIRC they would save two 
bytes per message in an already aggressively compressed high volume feed. I did 
some testing using the CME feed a while back and found that bit groups could 
contribute to a higher compression rate, but those tests were incomplete and 
would need to be rerun, possibly jointly with the CME. I'd guess that the Eurex 
EBS feed could benefit as well.

Bitgroups can also be emulated with uInt fields.


The 2.0 scope has virtually no empirical backing.

- map operator
This feature has been discussed on several different occasions, Kevin brought 
it up in conjunction with OTC instruments, a similar feature is used in one of 
the Xetra feeds, ...

Furthermore, besides not having a solid use case, we're not in full agreement 
on how the map operator should be implemented.

There is no convenient emulation available.

- Pmap placement
Dan and I did some back of the envelop calculations a few years back when we 
discussed the indivisibility problem with pmaps (you consume a byte as soon as 
you use a bit). The original idea was to use the extra bits instead of just 
wasting them. We also discussed the possibility to only use the pmap for one 
bit values, two valued fields like booleans.

We should probably analyze one-bit and multi-bit scenarios separately.


So, 1.1.1, 1.2, 2.0 or some variant thereof?

Best,
Rolf


[You can unsubscribe from this discussion group by sending a message to 
mailto:[EMAIL PROTECTED]

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"Financial Information eXchange" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/FIX-Protocol?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to