- From: <hallam@zorch.w3.org>
- Date: Mon, 26 Feb 96 13:14:22 -0500
- To: Adam Jack <ajack@corp.micrognosis.com>
- Cc: hallam@zorch.w3.org, www-html@w3.org
>> There is also a deeper philosophical issue here. The Web works because it >> is decentralised and has no central registry. >> >I am not sure I agree -- they all registered with Internic or their >internal DNS system. At a time any browser knows how to lookup an >find the IP address for a domain name. Also routing tables are >propagated to allow requests and responses to be delivered. Actually this is not the case, there is very little which stops you from setting up your own DNS directory. Absolutely nothing in the protocols stops you doing this. The people in Cuba set up the .cu hierarchy with absolutley no INTERNIC support. There are conventions which would make it difficult for J random bozo to set up a DNS namespace put there is nothing in the protocols to enforce this. What I was suggesting essentially boils down to reusing the exisiting DNs infrastructure. Actually it is slightly wider since if URNs were ever to appear there might be a direct route to stating an ISO template - ISO:13042.23 or whatever. >Additionally - are not the likes of Yahoo and WebCrawler a form of >centralized registry? Sure - the WWW works without them but they are >best located by being in the registary. But anyone can set themselves up as a Web index. Yahoo and WebCrawler simply set themselves up. They didn't have to ask anyone first. DEC set up Alta-Vista recently and have taken over a big chunk of the market. The point is that HTTP does not give a special role out to any player. There is no "if you can't find anything go to www.yahoo.com" provision in HTTP. There is a level playing field and each user can decide for themselves what works best for them. >Could that not equally be done by setting a field name standard and then >stepping back? We could have names like : >WWW.Personal.Name.FirstName = Adam >WWW.Personal.Name.LastName = Jack I think that this would interact very poorly with Java and other mobile code projects. I would much rather minimise the impact of automated forms filling and leave the names alone. For example my housemate uses the name fields in a form to pass arround pieces of code which amount to a LISP continuation. >Any 'centralized' registry could be a simple CGI script on a server. >These days firms jump on the bandwagon of offering 'public services' >-- and this need be no different. How many fields will the be >anyway? Hundreds? Never more than thousands -- and definately not >the Internic millions. I doubt that the load would be less than that involved running a mailing list and there are plenty of those which involve a full time job for someone - even with Majordomo etc. It would of course be open for someone to set up such a service and the URL they used to describe their template could be a link to a page describing their service and point at the CGI script which maintains the registry. The point is that there is no reason to build one model of naming things into the system. There is a tension here between people like myself who might only want to store 6 items and exmplicitly refuse to deal with schemes incorporating social security numbers and the like and vast, comprehensive schemes such as you describe. I suspect that inbetween these there is a place for an ISO like spec with 2000 odd carefully chosen and argued fields. One point I should have made, I made this in my similar proposal for HTML maths extensions the URI of the template need not be simply a name. It could be a location from which information about the resource might be fetched. For example a HTMl document describing each field. Alternatively it might be a piece of Java which would automatically generate the data - although such schemes would be very dicey and require a lot of thought which is why I didn't bring them up in the first place. Phill
Received on Monday, 26 February 1996 13:14:30 UTC