Theory: 'necessary' vs 'contingent' webpages
Jorn Barger
terms of 'necessary' and 'contingent' webpages.
These terms are from philosophy, and 'contingent' is more or
less synonymous with 'arbitrary' or 'accidental' or 'random'.
I'd suggest that the 'semantic web' can _only_ be built via
a backbone of necessary webpages on every topic, each one
handcrafted by an individual enthusiast.
The main way you make a page less random is by spending a few
hours doing web-searches to see what related pages already
exist, and designing your own page to *link* all the best of
them, in the most usable manner possible. [2]
A necessary webpage should resemble a FAQ, in providing quick
answers to all the most common questions on the topic.
As a rule of thumb, necessary webpages should probably offer
a timeline for the topic-- almost every topic has some aspects
that can be fitted neatly into the timeline-format. (My recent
exploration of victorianweb.org has made me reflect on the
difference between a timeline of someone's life, and a short
biography-- because VWeb usually offers both. I think bios
are necessarily contingent, because writing prose paragraphs
requires the author to fill in and smooth the transitions,
which ends up expressing much more of the author's own
value judgments. So necessary pages should probably favor
timelines.)
There's a very _bad_ convention in web-design, to link just to
the _homepage_ of a related website, and consider your job
done, but this is not acceptable by necessary-webdesign
standards-- every individual subpage on that site should be
linked separately, if it has useful content. The necessary
page thus becomes a universal table-of-contents to all other
sites on the topic. [3]
It's to be expected that this will mean hundreds of links on
a single page-- this is no problem if they're handled as one-
word 'text buttons' instead of long, meaningless repetitions
of the site's full title and description. [4]
'Linkrot' becomes a serious problem with hundreds of links,
but until authors have good tools for minimising linkrot,
I think the near-term solution is for authors of necessary
pages to use something like my 'open web content license'.
[5] This allows (encourages!) other authors to copy the
whole page, and update the links, so long as they clearly
declare the extent of their copying, and link the original
page.
I'm not sure if a short webpage on any topic could really
be considered 'necessary', but neither should it be 200k
of text and links. My own experience is that the size is
determined by the web-research phase: topics that are
already the subject of many pages may require a longer
necessary-page to integrate all those links... or if there
are already good necessary-pages treating various
subtopics it may be possible to link them instead of
'building them in'.
Some examples:
author-page: http://www.robotwisdom.com/jorn/pullman.html
website-page: http://www.robotwisdom.com/sites/salon.com
history: http://www.robotwisdom.com/science/classical/greece1.html
robots: http://www.robotwisdom.com/ai/robots.html
footnotes:
[1] content-centered design: http://www.robotwisdom.com/web/
[2] resource-pages: http://www.robotwisdom.com/we/resources.html
[3] one-layer design: http://www.deja.com/=dnc/getdoc.xp?AN=497230674
[4] text-buttons: http://www.robotwisdom.com/web/pragmatics.html
[5] content license: http://www.robotwisdom.com/web/license.html
Arjun Ray
jo...@enteract.com (Jorn Barger) wrote:
| 'Linkrot' becomes a serious problem with hundreds of links,
Such as references in the old "Dejanews Classic" style:
| [3] one-layer design: http://www.deja.com/=dnc/getdoc.xp?AN=497230674
This works, thanks to Google having done something about it. However,
this URL is just a link to a stepping stone. The redirected link is
compatible with the newer method of referencing articles in the Usenet
archive by Message-ID:
http://groups.google.com/groups?selm=an_497230674
IOW, given a Dejanews reference, Google will synthesize an ersatz
message-id from the old Dejanews internal article number. This isn't
100% foolproof, but pretty close. For first pass conversions of URLs,
it's simple enough to be automated. With some more effort, such a
program could use the translated version to retrieve the article, parse
the header for the Message-ID, and construct the prefered format:
http://groups.google.com/groups?selm=1duft9z.1si...@207-229-150-57.d.enteract.com
[f'ups set]
Jorn Barger
Nowadays Google prefers:
news:<16e613ec.02081...@posting.google.com>
or
news:16e613ec.02081...@posting.google.com
which it renders as a link to the quoted article.
> | 'Linkrot' becomes a serious problem with hundreds of links,
> Such as references in the old "Dejanews Classic" style:
> | [3] one-layer design: http://www.deja.com/=dnc/getdoc.xp?AN=497230674
> This works, thanks to Google having done something about it.
If it works, it ain't rotten. (Duh!)
> However, this URL is just a link to a stepping stone. The redirected
> link is compatible with the newer method of referencing articles in
> the Usenet archive by Message-ID:
> http://groups.google.com/groups?selm=an_497230674
Arjun has quietly trimmed an assignment to shorten his alternate:
http://groups.google.com/groups?oi=djq&selm=an_497230674
^^^^^^
I think this used to be required but I'm happy to see it's not now.
(One might worry that it will be again, someday.)
> IOW, given a Dejanews reference, Google will synthesize an ersatz
> message-id from the old Dejanews internal article number. This isn't
> 100% foolproof, but pretty close. For first pass conversions of URLs,
> it's simple enough to be automated. With some more effort, such a
> program could use the translated version to retrieve the article, parse
> the header for the Message-ID, and construct the prefered format:
>
> http://groups.google.com/groups?selm=1duft9z.1si...@207-229-150-57.d.enteract.com
[I will disguise my ps as a sig so Arjun will miss it:]
--
(Arjun suffers from a sadistic personality disorder, and enjoys nothing
more than flaming people for non-optimal markup. One theory is that
he suffered permanent psychological damage when he accidentally locked
http into "referer" for "referrer"-- note his 'prefered' above.)