- From: Mark Madsen <msm@ansa.co.uk>
- Date: Mon, 3 Jul 95 14:39:43 BST
- To: leslie@bunyip.com (Leslie Daigle)
- Cc: ura-bunyip@bunyip.com, uri@bunyip.com
Leslie Daigle writes: > Sorry for stepping into this thread a little late -- my mailbox > turned into a listserv war-zone lately, and careful excavation has > taken a while... I keep feeling sympathy for anyone on this list who went on holiday 3 weeks back and came back to find all the URN-related messages to sift through :-) > My understanding of the thread is that the proposal is to build > URC-searching with URCs themselves, defining a semantics for > handling URC templates as queries. That's mainly how I see it, with the important proviso that it's not really a proposal: more a pointer to the fact that the URC proposed spec actually allows this anyway. "Whatever is not forbidden is permitted." (Quiz: Who said that, and what were they wearing at the time? Answers on a postcard, please, but not to me. :-) The reasoning behind the original posting is as follows: if the spec allows URCs to be constructed that are basically objects in the strong sense of the term, can people not build systems which make use of this feature? In that case, why not attempt to foresee some of the issues and pitfalls that will arise? > Architecturally, I can not say I thnk this is the way to go. If we > do URC's right, they will represent characteristics of information > resources, and, yes, those representations are an integral part of > effective and efficient searching for resources. However, if I have > correctly understood it, the URC spec (or a closely-related one) > will also incorporate standards for how searches are handled. I completely agree that it is necessary to get URCs right. Architecturally speaking, the best guidance we have is the knowledge that systems consisting of distributed objects can be well understood, whereas without objects they are are muddy swamp. I am not saying that the data representations possible with URCs should be bent in some way so as to allow searching or retrieval more efficiently (especially given all the sins committed with efficiency as a justification). Search and retrieval are not even the only things that could conceivably be required of a URC object. I am only certain that soon someone will want/need to do something with URCs that I cannot visualise right now (mobility, multimedia, megalomania...?). Experience shows that extensibility is well served by use of object based designs. In a nutshell, am saying that the spec should not be made into a straitjacket that will not permit scenerios like the one I originally proposed. I also agree with your understanding that the URC spec will ultimately incorporate searching-related standards. So far, from my reading of the URC spec, it seems as if this is concerned with specifying a minimal guaranteed set of query primitives that can be exploited by search agents. > I think this is premature because URCs were constructed after > studying the task of _representation_ (which is presumably > understood), not search and retrieval (which isn't -- we don't know > all tasks which people and/or programs will need to have supported > by searching through URCs), and the examples that were cited in the > thread were for more than just search primitives. I could quote you a few piles of papers on metadata representations, well no, maybe that's an exaggeration, say one pile. Some of them are highly profound, some may be trivial. One of mine has references to most of the (other :-) good ones: Madsen, Fogg & Ruggles, "Metadata Systems: Integrative Information Technologies", Libri 44(3)237-257 (1994). The example I gave was geared to showing that there were reasonable, interesting, and complex things that one could quite legitimately ask for that were (i) not exlicity disallowed by the spec (ii) easy to fit within the letter of the spec, (iii) technologically possible with existing tools and languages like Tcl, Obliq, and Java. Any standard that excludes reasonable use-demands will not long be adhered to. > [Mark Madsen wrote:] > > > I'm carrying the quoted example below to maintain the context for the > > discussion. (Please note, I did not write this bit. My guess is it was Terry Allen or Ron Daniel) > > > asks whether we could use processing instructions to convey > > > that info instead. Here's the example: (My original example, severely pruned) > > > > <!doctype urc SYSTEM "urn:x-dns:uri.ansa.co.uk:method-dtd-7"> > > > > <urc> > > > > <author method="m1">Smith, F.</author> > > > > <subject method="m2">Cats</subject> > > > > <URL></URL> > > > > <results> > > > > (Initially empty, this container holds the results of the > > > > searches.) > > > > </results> > > > > <methods> > > > > <m1 lang="niceScript"> > > > > (A script written in the niceScript(TM) language that > > [snip] > > > I would prefer to see all this defined in a way that didn't depend on > > a particular language. (I am perfectly happy to see SGML used as the > > specification language for URC syntax.) > I urge you to go back and check out the URA Internet-Draft. The > structure of the example is not at all dissimilar to the kinds of > things we've been talking about for URAs. We haven't suggested an > implementation methodology at this point, but the test URAs we've > been playing with (in Silk) have been written in tcl and have a > similar structure. I shall go back and have another look at URAs. I don't have a problem with URAs doing similar things to what I described for URCs, myself, because ultimately every computational object on the WWW will want/need to be able to do things like this. If URCs and URAs are so similar, then there may be some point in merging the 2 concepts into one at the top level and giving them each their own subclass descriptions. > By contrast, though, URAs: > a) are not for searching only (although that's primarily what > we've done with them in Silk) As I explained above, URCs are not for searching either, since the method embedding I used contains computationally complete schemes. > b) take a step back to _describe_ the URA's data components (I > believe it was the "what" section in the paper), or the > attributes in the example above. This means that the URA > can encapsulate a search for subjects, which can be instantiate > to be a search for cats... Likewise, the example I cited could easily have been constructed so as to instantiate an interface to a search object. But the example would have been even longer. > I'll stop here, because I'm on a beastly slow link and generating > semantically parsable sentences is quite a challenge. I'd be happy > to continue talking about these things if we think there is > commonality between what is being described here and what we've > identified in URAs. There is commonality between what I originally described (URCs as autonomous agents) and any other distributed object systems. The message I am getting from your comments is that we all want to move the Web over to a distributed object model. That's great, because here we feel strongly about such things: see <URL:http://www.ansa.co.uk/phase3-activities/> I'd consider it worthwhile to prolong this discussion so as to learn better how to do that. > Leslie. Regards, Mark. -- ________________________________________________________________________ Mark Madsen: <msm@ansa.co.uk> <URL:http://www.ansa.co.uk/Staff/msm.html> Information Services Framework, The ANSA Project, APM Ltd., Castle Park, Cambridge CB3 0RD, U.K. <URL:http://www.ansa.co.uk/>; <apm@ansa.co.uk> Voice: +44-1223-568934; Reception: +44-1223-515010; Fax: +44-1223-359779
Received on Monday, 3 July 1995 09:40:14 UTC