From: tboyle@NoSMAProsehill.net (Todd Boyle)
Newsgroups: alt.accounting
Subject: Dayfiles architecture for webledgers on freenet
Date: Thu, 25 May 2000 19:34:50 GMT

On Wed, 24 May 2000, tboyle@NoSMAProsehill.net (Todd Boyle) wrote:

>http://freenet.sourceforge.net/ 

A webledger could be implemented immediately on FreeNet networks, based
on a dayfiles architecture.  Nothing preventing this, that I can see.

CPAs have been un-learning some of the lessons of the modem era. If 
network speeds start to *really* break out to higher speeds, it will 
drive accounting architectures in unpredictable directions.

George Guilder and Nicholas Negroponte were way ahead in playing with 
these ideas.  People have said, "disk prices will collapse!"  and even 
RAM would be unnecessary. You could access the single, authentic 
copy of the data for anything, faster over the network than you could pull 
it across the PCI bus.  

If speeds were infinite, all accounting systems could become 
serial-single-user systems.  You could download the entire database 
for each data entry.  When any remote person clicked on POST, they 
could check out the entire accounting database from its "home", 
download it, post a transaction, and send it back.

One approach that is feasable today is an accounting database 
consisting of 365 day-files sitting on a host computer.  Remote users 
could still have replicas of all data, and they would get the level of 
performance they are accustomed to within desktop software.  

When you need to post a change, you would send a lock request to the 
server, and begin working. You might send a checksum or CRC to the 
server,to ensure the dayfile has not changed before you begin working.

A thread running in the background would see if you have the latest 
version, and fetch it if not.  It would lock the dayfile, on the 
server.  The server app would be very simple. It could be done by a 
highschool kid with cgi. This is little more than an FTP server.

After finishing work, you would encrypt the revised dayfile, and 
upload it to the server.  Optionally, the server might propagate it to 
everybody else. 

The definition of an accounting transaction is a balanced set of rows, 
containing 2 or more debits and credits occurring at one point in 
time.  Even sales or purchases journals can be represented within a 
flat double-entry table with good effect; many of the leading small 
business accounting systems have straightforward double-entry tables 
for sales, purchases, payroll, etc.

Consistent timestamping would be important in this architecture, 
otherwise the client wouldn't know which dayfile to pull, particularly 
in multiple timezone situations, around the system midnight time.

This could be mitigated by putting a standard timeserver app on all 
the remotes.  Accurate timestamping would be a key to larger scale 
architectures based on Day-Hour files, or even Day-Minute files.  
Sooner or later you're going to run into Timeservers in e-commerce and 
digital cash anyway.  Why not hardwire the whole thing with 
timestamps. Timestamps potentially can have some extremely valuable 
internal control benefits.   Within ecommerce, timestamps have 
important nonrepudiation benefits, 


* Todd F. Boyle CPA    http://www.GLDialtone.com/
* tboyle@rosehill.net    Kirkland WA    (425) 827-3107
* XML accounting, WebLedgers, ASPs, GL dialtone, whatever it takes