This documentation needs to be thoroughly reworked to give the system a more suitable distributed design. Rather than having a centralised repository for managing URL and Query IDs these will be administered locally (as 32-bit IDs) and inter-host communication will use the string values (i.e. the URL or the Query text) as keys. Use of MD5 checksums for partitioning work will be appropriate. Maybe some clever DNS-based management system too.
John 07:14, 8 August 2011 (UTC)
John 07:24, 8 August 2011 (UTC)
DNS zones for query.cweb.blackbrick.com and url.cweb.blackbrick.com could be configured, and entries added for MD5 sum prefixes, for example a1.query.cweb.blackbrick.com and 054.url.cweb.blackbrick.com. The DNS names will resolve to a set of IP addresses. Those IP addresses can then be contacted via HTTP with IP-address host names, e.g. http://188.8.131.52/blackbrick-cweb/host-info, which will respond with the Cweb implementation's version number (I'm thinking a datestamp with serial, e.g. 2011-08-11-1, allowing for up to 9 releases per day and supporting string-sorts), with the DNS name the service should be contacted on (e.g. jj5.user.cweb.blackbrick.com) and whether or not HTTPS is supported.
John 07:31, 8 August 2011 (UTC)