Hi Eddy, You are right. That's why I have introduced a new data warehousing scheme in the R-Quant framework. I will let you know when I update the documentation on the web. You are also absolutely right saying that I can not ask root team to serve financial community in general and Fortis bank particularly. That's why I am switching from ROOT dependent implementations to virtual interfaces to data sources APIs (Bloomberg, Reuters, TAQ, etc) and data servers (Oracle, SQL Server, etc.). If we are talking about tick recording and further analysis, then the most efficient implementation is a fixed record size binary stream. Although this approach might look too simple, there are not many objects with variable sizes in the financial data warehousing. It's all about quotes and trades. All the other backoffice objects can be easily stored with standard RDBMS tools. Well, let's put it like this. My hobby, apart from physics, girls, billiard and champagne, is rquant, which is suppose to be a general framework for financial data analysis. I understand that every large financial institution, including Rentec and Fortis, has a gang of programmers/analysts which can write and rewrite trees day and night, although this is more about Exel tables and SQL requests to Oracle in reality. But it is quite hard to explain "one write many reads" issue to an average customer of the framework, which suppose to be general enough to serve people doing different things (including university researchers in finance for example). On the other hand in my data warehouse each quote/trade has a field with a confidence level/telling how the data were acquired (or approximated for non-business days). If I get a correction, I would definitely like to be able to change this field. October 97 TAQ database came for example with a number of consequent corrections and recommendation for data crosschecking with different vendors. These crash data are very interesting for pattern recognition/predictions and I am sure you want to get them right. It means you may get confirmation about several concrete entries after quite long time when your datawarehouse has been already filled with gigabytes of other trades. > I have seen you raise again the point of manpower on the ROOT project but in the > same breath you mention inefficient management of these large-scale experiments. > So what do you want ? These days, being an ordinary taxpayer, I want people to spend a part of the tax money intended for those large-scale experiments in somewhat more efficient way. PS. I keep the door open to change my opinion in case I am back to science and experimenting :) > An easy way to put your mark on the project is to open > a can with programmers at Fortis-bank and implement your vision. Because > let's face it at the moment you and I have a free ride using the ROOT team. Well, the neural networking and stochastic optimization packages are free at the www.smartquant.com/neural.html and R-Quant classes/sources are there too. Everybody is welcome to take a free ride and use it. On the other hand I think it is quite good for developers if people use the soft, report bugs and send requests for improvements. I am sorry for discussing perhaps a non-related topics on the rootalk. Eddy, I am setting a mailing list software on the smartquant.com, so you are welcome for flaming very soon :) Regards, Anton http://www.smartquant.com
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:38 MET