I hate Netscape's bookmarks, so instead I have three pages on my local
hard drive:
1) a startup page with my most frequent links (incl many local
directories, for html hacking)
2) a secondary page with infrequent bookmarks
3) a search page with a form for each of my fave search-engines, with
the settings I favor.
These load pretty instantaneously, and link to each other so I can jump
between them, and I also used a utility to program them onto my Netscape
button-bar.
The search page was the trickiest, so I'll include it below. It's
pretty self-explanatory except for Rankdex, which seems a lot more
effective than Yahoo if I just want a decent page devoted to some random
topic (recent successes: Van Gogh, Debbie Harry, HTML tables). The IMDB
form is for movie titles only, which is what I usually go after.
(I don't actually know what all the hidden variables stand for-- many
may be unnecessary.)
- Altavista WWWeb
- Rankdex
- Yahoo
- IMDB
- Old Deja Archive by score
- Current Deja Archive by date
- DejaNews Author Profile
===
<FORM method=GET
action="http://www.altavista.digital.com/cgi-bin/query">
<INPUT NAME=q size=40 maxlength=800 wrap=virtual value="">
<INPUT TYPE=hidden NAME=act VALUE=search>
<INPUT TYPE=hidden NAME=pg VALUE=q>
<INPUT TYPE=hidden NAME=text VALUE=yes>
<INPUT TYPE=hidden NAME=what VALUE=web>
<INPUT TYPE=hidden NAME=kl VALUE=en>
<input type="submit" value="Altavista WWWeb">
</form>
<FORM action="http://rankdex.gari.com/cgi-bin/query" method="get">
<INPUT name="query", size=40>
<INPUT type="submit" value="Rankdex"></FORM>
<form action="http://search.yahoo.com/bin/search">
<input size=40 name=p>
<input type=submit value=Yahoo></form>
<FORM ACTION="http://us.imdb.com/M/title-substring" METHOD="GET">
<INPUT NAME="title" SIZE=40><INPUT TYPE="submit" VALUE="IMDB">
<INPUT TYPE=hidden NAME="tv" VALUE="off"></FORM>
<form action="http://search.dejanews.com/dnquery.xp" method=post>
<input type="text" name="QRY" size="40">
<input type="hidden" name="ST" value="PS">
<input type="submit" value="Old Deja Archive by score">
<input type="hidden" name="defaultOp" value="AND">
<input type="hidden" name="svcclass" value="dnold">
<input type="hidden" name="maxhits" value="50">
<input type="hidden" name="format" value="terse">
<input type="hidden" name="showsort" value="score">
</form>
<form action="http://search.dejanews.com/dnquery.xp" method=post>
<input type="text" name="QRY" size="40">
<input type="hidden" name="ST" value="PS">
<input type="submit" value="Current Deja Archive by date">
<input type="hidden" name="defaultOp" value="AND">
<input type="hidden" name="svcclass" value="dncurrent">
<input type="hidden" name="maxhits" value="100">
<input type="hidden" name="format" value="terse">
<input type="hidden" name="showsort" value="date">
</form>
<form action="http://search.dejanews.com/profile.xp" method=post>
<input type="text" name="author" size="40">
<input type="submit" value="DejaNews Author Profile">
</form>
===
j
--
I EDIT THE NET: <URL:http://www.mcs.net/~jorn/html/weblogs/weblog.html>
"The PhD system is the real root of the evil of academic snobbery.
People who have PhDs consider themselves a priesthood." --Freeman Dyson
> 1) a startup page with my most frequent links (incl many local
> directories, for html hacking)
> 2) a secondary page with infrequent bookmarks
> 3) a search page with a form for each of my fave search-engines, with
> the settings I favor.
I experimented with a local search page, on my server
rather than on my hard drive. It had a search box
for the Mamma metasearcher and a directory front-end to
EZ-Connect. It's orphaned now, but the sad remnants may be
seen at http://www.teleport.com/~lensman/yarp.htm
My home page consists of three links to my own sites and
three to other favorites, gussied up with some graphics and
formatting. That's all! Probably less a purist statement
than just my lame attempt at minimalist cool, though.
>Rankdex, which seems a lot more
>effective than Yahoo if I just want a decent page devoted to
>some random topic
If you like Rankdex's laser pinpointing of just a page or
three, I recommend a look at google.stanford.edu which uses
a similar but expanded quality algorithm and a much larger
index.
--
Chris Mitchell
The Searcher's Road Less Travelled
http://www.teleport.com/~lensman/roadless.htm
I'm baffled, and have to guess that you misunderstand Rankdex. Google
gives me lots of results for "van gogh", all with better than 99%
rating. The technical paper is so wordy I couldn't tell quickly where
the rating comes from, but I don't see any sign it was rankdex-like.
To re-explain the very simple principle of Rankdex:
- Linking to a page means you like it. Counting these votes gives a
rating for a page.
- Anchortext for the link usually summarizes the content. When your
query is straightforward, like "Debbie Harry" this becomes a highly
reliable way of finding a good Debbie Harry fanpage.
True, Google typically returns a list of pages, but Rankdex
can return as many as ten also. In fact, "van gogh" returns
ten. Rankdex does give more discrete rating values in this
case, but part of this must be due to the fact that Google's
index is 24 million documents and Rankdex's can't be more
than a million or so, if that.
> To re-explain the very simple principle of Rankdex:
>
> - Linking to a page means you like it. Counting these votes gives a
> rating for a page.
>
> - Anchortext for the link usually summarizes the content. When your
> query is straightforward, like "Debbie Harry" this becomes a highly
> reliable way of finding a good Debbie Harry fanpage.
Google uses the exact same techniques - identical! It also
adds the following: a link from a page which itself has a
high rating is weighted higher than a link from a page
which has a low rating.
So both indexes derive quality from link popularity and
anchortext, with Google adding an extra level of recursion.
--
Chris Mitchell
len...@teleport.com
The Searcher's Road Less Travelled
Along the way, then, Google replaces the simple vote-count with an
opaque percentage, adds an unhelpful bar-graph for each response,
etc etc etc.
Rankdex might be slightly better if there were a lot more pages,
but what I mostly want it for isn't exhaustiveness (for which I
use AV) but simple straightforward popularity votes.
A typical Rankdex response will show one page with several hundred
votes, and the next with half that... making the choice very clear.
Google instead gives me a long list of 90%+ ratings, making the
choice very unclear.
Also, Google's search-page is bogged down with ridiculous amounts
of history/theory/acknowledgements, making it load much slower,
and menus that aren't really relevant for me. And the search
results include many ringers that presumably derive from the
'recursive' level, which I just don't see the point of.
j
ps: Interesting Rankdex search-pattern: "Click Here"
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/ Now offering spam-free web-based newsreading
You also raise some good issues about Google's recursive
weighting. I will go back and wade through their
technical papers sometime. It does worry me that if the
"popularity rating" technique becomes common, it will
be more difficult for high-quality newer pages to get
exposure. Hopefully there will always be hand-selected
directories like Yahoo and InfoSeek Guide to counter
such a trend.
>ps: Interesting Rankdex search-pattern: "Click Here"
Good one!!
--
Chris
[I've had this conversation in no less than 3 places in the past 2 weeks.]
From <http://xent.ics.uci.edu/FoRK-archive/mar98/0311.html>:
What would be great is a page that is just one large form, with one text
input box where you could enter any text you want, the contents going to any
number of most-often used services depending on which submit box you hit --
so you could have, for instance, submits for altavista, hotbot, yahoo
search, deja news, the m-w dictionary and thesaurus, etc. filling most of
the page with only one text-entry field. This would require a local cgi app
to interpret the form and send the response, I think. Or maybe better yet,
if you're doing the programming then eliminate all the submit boxes and have
the cgi interpret the submit using a language as part of the text itself (so
you don't have to use the mouse), like "dn" for deja news, "av" for alta
vista, etc...
I have an early version of my most-often used starting points (with lots of
seperate forms, unfortunately) at
<http://dsl.org/m/doc/comp/info/search.html>.
--
Michael Stutz . http://dsl.org/m/ . copyright disclaimer etc
st...@dsl.org : finger for pgp : http://dsl.org/copyleft/
Do you mean like http://www.dnp.fmph.uniba.sk/~garabik/search.html
________________________________________________________________________
Stud. Scient Knud Haugaard Soerensen homepage:
http://www.mi.aau.dk/~khs/
Departments of Mathematical Science University of Aarhus, Denmark
Freak Technology: News on Technology http://sunsite.auc.dk/FreakTech/
Aeiwi: The Future of Web Indexing http://www.aeiwi.com/
> Do you mean like http://www.dnp.fmph.uniba.sk/~garabik/search.html
That's it! But (personal preference) I would put the text box at the top of
the page instead of the bottom.
Is the source of this cgi available?
1998年3月31日火曜日16時00分00秒 UTC+8 Jorn Barger:
> Chris Mitchell <len...@teleport.com> wrote:
> > If you like Rankdex's laser pinpointing of just a page or
> > three, I recommend a look at google.stanford.edu which uses
> > a similar but expanded quality algorithm and a much larger
> > index.
>
> I'm baffled, and have to guess that you misunderstand Rankdex. Google
> gives me lots of results for "van gogh", all with better than 99%
> rating. The technical paper is so wordy I couldn't tell quickly where
> the rating comes from, but I don't see any sign it was rankdex-like.
>
> To re-explain the very simple principle of Rankdex:
>
> - Linking to a page means you like it. Counting these votes gives a
> rating for a page.
>
> - Anchortext for the link usually summarizes the content. When your
> query is straightforward, like "Debbie Harry" this becomes a highly
> reliable way of finding a good Debbie Harry fanpage.
>
>