Introduction

Twiki can be the best vehicle to break the tradition of server-side database approach, and therefore to stimulate the progress of user-centered content management philosophy.

Below is the example.

The Problem

Now the library catalogs are built according to database philosophy. One information object (resource) is described by one record in the database. The harvesting machine, replying for user's request, displays the list of records.

The catalogs written in the webpages inherit this traditional database logic: one object has to be described in one separate webpage - say web document. The searching engine displays the list of web documents.

So, one has to keep thousants of separate web documents in the archive of metadata. Filling them is time consuming, because you have to create separate page for each record (10-20% of time is consumed only for saving one record and opening new one).

Using XML/RDF even worsens situation - look for example at METS specification - there is much more code than content! Processing such bunch has to consume so much machine power.

But in fact, a record is the unit of description (metadata), not webpage!

For about three years I'm looking for the reverse solution: to write many records in one web page, and harvest for the single records not web documents! Something like RSS. For information processing from the client side I tried to apply DTML manipulation of DOM structure, to parse XML tree etc. All appeared to be too complex or too restrictive for average user demanding loose structure of the record.

Finally I came to Wikis and realized they are nearest to the proper and simple solution because they are DOM-sensitive. I tested Zwiki for well working comments, tried to modify its issuetracker, tested Purple Numbers for explicit marking the paragraphs...

After all my tests Twiki appeared to be the best because:

• it writes the topics in txt files, not database,
• it already has some usful DOM-processing, like TOC and Side Show Plugin; so the community seems to be open for the idea of processing the sections of the webpage.
• it has searching facility sophisticated enough to retrieve the fragments of webpages, and easy retrieval of DOM section seems to be possible.

I concluded that Twiki can be the best vehicle to break the tradition of server-side database approach, and therefore to stimulate the progress of user-centered content management philosophy. It is easy to see that many participants of Twiki community have similar ideas.

Above is the extension added on 08 Feb 2004 -- AndrzejGoralczyk

Question

Each topic in my library contains one or more sections, beginning with

---+++ <dc:title> The Title of the Article

and ending with, say,

-----

which displays a line. "The Title of the Article" is an example, of course.

1) How can I extract all the sections in the web, containing a word or phrase, say, "the Article"?

I tried to use formatting search, however I'm short in REG EXP.

Is it possible to write a search form for such an action?

2) I tried also to use INCLUDE a page "full_link_to_the_topic#The Title". This includes all the topic, and ignores what is after hash. How can write a pattern in order to include only a single section?

In fact, my idea is to combine a formatted search with INCLUDE, but I think, that the phrase searched should include wildcards and some characters indicating the beginning and the end of the section. Secondly, I don't know how to input an INCLUDE directive into format in SEARCH command.

Environment

TWiki version:	TWikiRelease01Feb2003
TWiki plugins:	DefaultPlugin, InterwikiPlugin
Server OS:
Web server:
Perl version:
Client OS:
Web Browser:	Mozilla 1.5.

-- AndrzejGoralczyk - 03 Feb 2004

Answer

AndrzejGoralczyk - what you're after is the ability to include implicitly named sections. TWiki does not have this capability. named include sections does provides similar functionality (not currently merged as part of TWiki due to it being merged last year in a way breaking large amounts of other people's content - it was backed out as a result (amongst other reasons)). If you want named include sections however - which might be what you want, please contact me.

Peter Masiar's modified comment plugin uses implicitly named sections to define templates, so if you're a perl coder, take a look at his code, given it already does much of what you're after. You can also look at the SlideShowPlugin for this sort of functionality, since again it looks at the DOM rather than imposed extra explicit sectioning.

-- MS - 06 Feb 2004

May be Sandbox.SearchBookTitles can help?

-- PeterThoeny - 08 Feb 2004

Thank You, MichaelSparks and PeterThoeny. I extended this topic above the Question section, in order to better explain my approach, based on 3-years experience in building web gateway to the scientific libraries and databases - a kind of web library for public use, optimized for modem users :). Having no proper solution I interrupted this project, because it was no sense to continue it in the simplest way, extremely time-consuming. However, it is still the only such project in Poland. Now I think that Twiki could be the platform for new version. It is necessary to add, that new opprotunity for promoting and developing such solutions will appear soon - a big project of Polish Knowlegde Management Forum (I am one of the leaders in this project).

Unfortunately, I'm power-user only, not a coder. Coming to the point - It would be great to have something like that:

%SEARCH{ search="phrase"... format="   In $topic see: $sectiontext" }%

Peter, I am not familiar with Twiki community customs, and I don't know if such topic shouldn't be located in Codev section. If yes, please move it.

-- AndrzejGoralczyk - 08 Feb 2004