Source

oclegacy / draft.txt

December 2003 report on the status of OverChat
==============================================
by David Duncan Ross Palmer

For those is may concern, I have contructed a report on the status of OverChat and listed any plans I migh have in mind for the new year.

1. The website
2. Protocol issues
3. OverChat and LLS
4. Open Source/Free Software
5. Fee/pricing model
6. Next release
7. News on Dagsylad
8. New Year's message

1. The website
==============
After another downtime period, the website has been restored.  Unfortunately, because I have other things to work on, business issues, housework and the like, sometimes I just have to write these problems on a big list marked "to do" on my wall.  The problem this time was a permissions problem on the SQL server.  SQL is in use for the website only but will be used by OverChat in a couple of releases' time.  The problem was caused because we here at the WoodenAxe LAN (the network which currently hosts OverChat's primary server and site), changed the domain from woodenaxe.de (which we decided not to register because of cost) to woodenaxe.net (which is registered and operational).  This meant that the SQL server did not recognise the machine hosting the website and would not allow it to access parts of the site stored in a database.

2. Protocol issues
==================
After careful consideration I have decided to reject Cristian Klein's idea for protocol extensions.  These seemed plausable but as he has not implemented them and I would have to implement them myself, I am much more favourable to a comment I once read on another web site about binary versus text protocols.  The limitations Cristi thinks are in text mode protocols aren't true, as the protocol can switch to passing binary chunks of data, either as uucode or binary proper at any point which both ends of the conversation can agree on.  The advantages of a text mode protocol are

1) Extendabillity, without the need for reserved fields
2) Easy debugging when things go wrong
3) Third-party extensions which don't break the official protocol
4) Thin-clients which do not wish to use the Blake protocol
5) Much faster and easier development (something I am sure some people would appreciate!)
6) No endian worries

Binary protocols are great for the core architecture of the internet but not for user space services, mail servers, web servers etc. _all_ use text-based protocols.

I thought for some time about having a new Blake protocol packet which would request a switch from binary to text mode but this effectively destroys point number four!  Also there is the closed binary client "Sarah" written by Dagsylad Enterprises which needs to continue to run.  So I think this solution is the best idea:

The binary protocol existing on port 3277 would remain in place.
The query server on 3279 will be phased out, as has long been the plan.
Port 3278 will be opened which will use the text mode protocol, this will be written from scratch.

No further development will be done to the Blake binary protocol unless a client developer desperately needs it.  At the moment there is only one (known) third party developer (Dagsylad).  If he requests a new feature, it will be added (if possible) but any further additions will go to the text-based protocol only.

Clients _must_ migrate to the text protocol and the binary protocol will eventually be phased out and port 3277 will be closed.

Please not that port numbers are the responsibillity of individual server administrators and are not part of the protocol design.  Numbers here do not imply mandatory port numbers.

3. OverChat and LLS
4. Open Source/Free Software
5. Fee/pricing model
6. Next release
7. News on Dagsylad
8. New Year's message
(Programming is the most important thing)


-- 
D. D. R. Palmer of Daybo Logic, "We enjoy solving your problems".
For phone or email details http://www.daybologic.co.uk/mailddrp
Warning: The sending address may change at any time, without any notice.