- From: Patrik Faltstrom <paf@bunyip.com>
- Date: Mon, 3 Jul 1995 17:44:52 -0400
- To: "Ronald E. Daniel" <rdaniel@acl.lanl.gov>, uri@bunyip.com
At 14.55 95-07-03, Ronald E. Daniel wrote: >Your comments please! My opinion is that this group should only come up with an _architecture_ for information retreival, not an implementation. This means that if for example the common indexing protocol is used for exchanging URC's between the whatever-is-storing- the-URCs, that protocol is defining the syntax of the messages that is sent between servers. This group, the URI-group, should not try to invent the protocols, storage formats etc. As soon as you have a description of the functionality of the URN/URA/URC/URLs I will implement this based on the template-format used in Whois++ (in fact I have already one implementation running). For example, is an URN mapping to one and only one URC (which might be mirrored) which might point to several documents by containing several URLs? Can one document have several URNs (one ISBN and one OID for example)? What is the syntax of a URN? How should each of these items be used and how to resolve a URN? That is the kind of architeture problems I want to have the answers for when trying to implement this stuff. We already know that there is several other implementations part from mine, and we have to in the future also to decide if we will go with one protocol or several for each step in the resolution process. IF we are going to go with one only in some or all of the steps, we have to decide which to use, but not until we have seen some implementations, thanks. This group is setting up the rules, and the rules should not be "you must use DNS in this step of the process" or "you have to have HTML syntax there" or "this is done by using Whois++" or whatever. You should _only_ define the syntax and protocols used when it is _really_ essential, and I can so far only think of the syntax of the URN, or not even that, because the syntax of the URN is defined in each of the alternatives we have. So, my implementation, using Whois++ for the client-server communication and the common indexing protocol for the server-server communication is for example not using SGML syntax at all when storing the URCs. What format to use internally is an implementation issue. The architecture I am using for _implementing_ the URA/URN/URC/URL scheme is discussed in the ASID group, and it should stay there. Patrik
Received on Monday, 3 July 1995 17:44:33 UTC