- From: John Nissen <jn@tommy.demon.co.uk>
- Date: Thu, 26 Aug 1999 17:35:49 GMT
- To: jbrewer@w3.org
- Cc: w3c-wai-ig@w3.org, jn@tommy.demon.co.uk
Judy, I got a reply from Toby on chi-web, to which I've responded, see below. Cheers, John -- Forwarded message follows: >Date: Thu, 26 Aug 1999 16:51:56 GMT >From: jn@tommy.demon.co.uk (John Nissen) >Reply-To: jn@tommy.demon.co.uk >Message-Id: <52674@tommy.demon.co.uk> >To: CHI-WEB@ACM.ORG >Cc: jn@tommy.demon.co.uk >Subject: Re: Tree model integrity >Hello Toby, > >Thanks for your considered reply. > >In message <19990826015900.2688.qmail@hotmail.com> tobyhede@HOTMAIL.COM writes: > >>i think that it remains to be seen that all information can be represented >>in a tree/hierarchy. The tree system is a representation that stems from the >>last 20/30 years of file system hierarchies...nielsen, among others has >>pointed out that many new computer users have a great deal of trouble with >>the tree or hierarchical methodology. the tree is limiting as an information >>methodology, it disallows the entire network principle that makes the web >>such a valuable medium. > >One can have the web as a forest of trees. What I'm saying is that each >site can and should generally have a hierarchical structure, unless there >is a good reason otherwise. > >>a tree system permits only one classification for >>information...each 'unit/block' of information resides within one branch of >>the tree, but different users will have different classifications for the >>same information. using a strict tree structure your are in reality >>enforcing your classification upon them. > >I disagree here. Classification has little to do with structure. > >>Searching: >>i agree that having to page back through the search results in order to >>perform another search can be frustrating, but a page with 10,000 search >>results on it would be rather slow to load and difficult to navigate through >>coherently. a better way to resolve this issue is to have a 'refine search' >>input area at the top (and perhaps bottom) of the page. some of the better >>search engines do indeed use this technique. > >The frustration you mention is an example of what I'm getting at. >However I'm not expecting search engines to return all 10,000, but the >usual 40-60 or so. These are normally put on up to six pages or so, >but I propose that they should all go on one page. Then this particular >frustration is avoided. > >I slightly disagree about the refinement for two reasons. A refinement >generally uses the same form as the original search. Secondly, when I >abandon a search I want to abandon any refinements at the same time, so >I'd like Back just to take me out of the whole process in one click. > >>Back Buttons: >>there are two issues here. the browser's back button takes you back through >>the history of navigation actions. but, a great deal of information forms >>part of a logical stream in itself. especially with deep links forming an >>integral part of the web (see below), the need to link within a stream of >>information is vitally important. imagine an encyclopedia or dictionary that >>refused to let you turn to the previous or next pages, but would only let >>you return to the index or table of contents (or wherever it was you last >>were)...i think the art is in differentiating between the two types of back >>and forward activity. > >I believe that the encyclopedia and dictionary are exceptions in that >cross-linking is the norm - i.e. navigating items of information that >are at the same level, significance or importance. I'd have to accept >that Back would take me through all the items I'd visited - rather >frustrating. The only way around this would be to have the browser's >backward link unaffected by your navigation sideways links - and that >would require changing how browsers work! > >I'd like the normal case to be that Back button takes me up the >hierarchy of information on the site. > >>Site Entry: >>http://www.useit.com/alertbox/990725.html >>if a user needs to hit your site from the top in order to map your >>information, than the navigation system is probably not transparent enough. >>menu structures and page titles can provide instant user recognition of >>location within a larger information structure, plus provide the flexibility >>that a networked information structure demands. > >That's not what I'm getting at. There is little information at the top >level, and it mostly concerns what is at lower levels. So the user can >quickly go through the first level or two to get at the guts of the >information. To speed the process an index may be desirable, but I'd >put it at a higher level than the information that it is indexing. This >may seem counter-intuitive, but it keeps all links going down the >hierarchy, and this is important. (Exceptions are the cross-links from >one part of the index to another.) > >Cheers, > >John >-- >>>From: John Nissen <jn@tommy.demon.co.uk> >>>Reply-To: jn@tommy.demon.co.uk >>>To: CHI-WEB@ACM.ORG >>>Subject: Tree model integrity >>>Date: Wed, 25 Aug 1999 21:02:57 GMT >>> >>>Hello, >>> >>>A few weeks ago we had a thread about the importance of building >>>up a coherent mental model, so you know where you are when you >>>are navigating a site (but, if you get lost, you can quickly find >>>yourself again). >>> >>>Now, during your navigation of the web, the browser builds up a >>>tree of visitations. You are aware of this when you use the back >>>and forward buttons, and also when the colour of links change. >>> >>>I believe many problems that users have with web applications are >>>due to the applications messing up the browser's tree. Sites should >>>have all information in a hierarchy, so that the model of the site >>>maps onto the browser model. And sites should avoid "short cut" buttons, >>>instead leaving the user to use the browser to do the navigation. >>> >>>I claim: >>> >>>A. It is possible to present almost all information in a tree hierarchy, >>>where the nodes are pages and the branches are hypertext links. >>>(Help information, glossary of terms and dictionaries are exceptions, >>>since cross-linking is inevitable.) >>> >>>B. Most navigation buttons introduced to make it faster or simpler >>>for the user actually have the opposite effect, because they >>>subvert the browser's own controls. >>> >>>Some implications: >>> >>>1. Searching >>> >>>The search engine should present results on a single page. >>>It should not break the results into a number of pages and have >>>"next" and "previous" buttons. Instead the user should be allowed >>>to scroll. The browser back button will then take the user straight >>>back to where they started with the search form, no messing. >>>Search refinement should be from this same point. >>> >>>2. Back buttons >>> >>>The application should never supply its own back, forward >>>or home buttons. >>> >>>3. Internal links >>> >>>There should be no links between different points of the same >>>page. All links should point down the hierarchy. (Exceptions >>>are for cross-referencing in help files and dictionaries.) >>>Applications should never have "top of the page" buttons. >>> >>>4. Site entry >>> >>>Users should be encouraged to visit the site starting at the >>>top. Then the tree of the site will map directly onto the >>>browser tree as it creates it. >>> >>>5. Table of contents or index >>> >>>This should be on a separate page at a level above the contents >>>itself. >>> >>>6. References >>> >>>These should generally be in a page by themselves at the lowest >>>level. >-- >Access the word, access the world Tel/fax +44 181 742 3170/8715 >John Nissen Email to jn@tommy.demon.co.uk >Cloudworld Ltd., Chiswick, London, UK http://www.tommy.demon.co.uk > -- Access the word, access the world Tel/fax +44 181 742 3170/8715 John Nissen Email to jn@tommy.demon.co.uk Cloudworld Ltd., Chiswick, London, UK http://www.tommy.demon.co.uk
Received on Thursday, 26 August 1999 12:41:06 UTC