html-template-users

Re: [htmltmpl] RFC: Template Tag Attributes

[Pete P. <pet...@cy...>]

Bill Moseley wrote:
> On Thu, Jun 03, 2004 at 12:33:18PM -0500, Pete Prodoehl wrote:
> 
>>I've seen sites where I could read a word on a page, input it into the 
>>'site search' box, and get no results. This tells me that the word does 
>>not exist on the site (even though I can see it) or more likely that 
>>this is not really a 'site search' but perhaps a 'content search' or 
>>'article search' or whatever...
> 
> Of course, if the word is on every page on the site (a global header or
> footer) the word is useless as a search term for helping someone find
> something on the site.  Unless they are looking for everything. ;)

Well, with properly weighing, titles, <h1>, etc. being of more 
importance, this shouldn't be the case...


> There's also the rare chance that indexing the common text makes it
> harder to find content about something specific -- when that word might
> also be in the header/footer.  Ranking should take care of that, though.

Yup... ;)


> Excluding menus is also slightly helpful when common menu items might end up
> as search terms.

Yup...


> But, I would agree that in most cases it's better to index the content,
> and let the user worry about using good queries.  Anyway, how much
> common content do you want on every page on the site?

How much do *I* want on every page, or how much to the "powers that be" 
want on every page? ;)


Pete

Re: [htmltmpl] RFC: Template Tag Attributes

[Timm M. <tm...@ag...>]

At 09:02 AM 6/4/04 +1000, Mathew Robertson wrote:
<>
> > Google PageRank is very good at searching a broad sample of sites.  It's
> > not so good for individual sites.
>
>You are kidding right?  The algorithms the Google use, already take into 
>account common content on multiple pages.  OpenOffice.org use Google as 
>their own site specific search engine.  As do a number of sites.  The only 
>real problem with using Google is that they only spider the web every few 
>weeks, thus if you update more frequently than that, you may have a problem.

No, I'm not kidding.  It's not common content that Google has a problem 
which makes it less than perfect for individual sites.  The whole point 
behind PageRank is that you get a lot of sites linking to each other and 
are given rank based on the link count.  On a smaller sampling, there 
simply isn't enough data to be fed into PageRank to make a good rank.  Is 
it still useful?  Sure it is, but it's also not as good as other solutions 
for this particular problem.

Re: [htmltmpl] RFC: Template Tag Attributes

[Mathew R. <mat...@re...>]

> > > Google PageRank is very good at searching a broad sample of sites. =
 It's
> > > not so good for individual sites.
> >
> >You are kidding right?  The algorithms the Google use, already take =
into=20
> >account common content on multiple pages.  OpenOffice.org use Google =
as=20
> >their own site specific search engine.  As do a number of sites.  The =
only=20
> >real problem with using Google is that they only spider the web every =
few=20
> >weeks, thus if you update more frequently than that, you may have a =
problem.
>=20
> No, I'm not kidding.  It's not common content that Google has a =
problem=20
> which makes it less than perfect for individual sites.  The whole =
point=20
> behind PageRank is that you get a lot of sites linking to each other =
and=20
> are given rank based on the link count.  On a smaller sampling, there=20
> simply isn't enough data to be fed into PageRank to make a good rank.  =
Is=20
> it still useful?  Sure it is, but it's also not as good as other =
solutions=20
> for this particular problem.

PageRank is not the only ranking algorithm in use during Google searches =
-> this is particularily the case when a 'site:...' search is done.

As with most other search engines, you can submit your site/url to =
google so that it is spidered then next time google updates its database =
-> you dont _need_ to be linked to, to get ranked (although it helps).

Mathew

Re: [htmltmpl] RFC: Template Tag Attributes

[Mathew R. <mat...@re...>]

> > > Inevitably, there will be certain pages using TMPL_INCLUDE tags.  =
I imagine
> > > that most of these will contain data that will not want to be =
searched for,
> > > such as footers, and therefore my filter program can simply ignore
> > > them.  However, I don't feel safe in making the blanket assumption =
that
> > > /all/ included files don't need to be searchable.
> >
> >Now you've lost me.  There's lots of stuff in an HTML page that
> >shouldn't be searched for.  Stuff like headers and footers in =
includes
> >is just the tip of the ice-berg.  Why obsess over this?
>=20
> I want to give content authors more control over what portions of a=20
> document are searched for.

This is a common mistake that information creators think 'is a good =
thing'...  The web got popular for a number of reasons - one of them =
being "full text indexing of all content" (including =
headers/footers/etc).

The point is that, it is the user of the system that wants to find the =
information - not the author telling you what you can and cant search =
for.   Classic example -> books used to have (and still do) an index in =
the last couple of pages of the book, yet the user could never find what =
they were looking for; until the book made it onto CDROM at which point =
full-text-searching was possible.

-> Full text searching is a _much better_ solution to search problems =
than indexing on what YOU think is the information they want.

Mathew

PS.  This means, use a spider.. or even better use google via a =
"site:..." search.

Re: [htmltmpl] RFC: Template Tag Attributes

[Mark F. <mar...@ea...>]

From: "Pete Prodoehl" <pet...@cy...>
> I've seen sites where I could read a word on a page, input it into the
> 'site search' box, and get no results.

Some search/indexing tools exclude common and site-defined words. I used a
really nice local indexing tool called WebGlimpse[1] and it had this ability
(as well as fuzzy word matches).

> While you say that there is no useful information in headers/footers,
> that may be by your definition and decision of what goes in a header and
> footer. This can vary from person to person.

A good indexer would add weight to words that appear in headers/footers
which also appear in title, meta keyword/description, headings, and
expository content (paragraphs, list items, data terms and definitions).
Conversely, reduce weight for words appearing in headers/footers (or any
part of the document) that doesn't appear in outline/structural elements.
What I really hate is all words presented to me without any weighting of
relevance to the topic of the page (determined from its structual elements).

If it were me, I would index the entire site and work on the relevancy
algorithem. For example, perhaps the footer should be enclosed in a <div
class=footer> and the indexer programmed to omit such classes of divisions.

FWIW: I don't think Tim's attempt to get more flexible/enabling features
will go very far. I can't even get an attribute indicating that linefeeds
only for the readbility of <tmpl> tags not remain when the tags are removed
by h::t.  It's like a festering wound for me. :)

[1] http://webglimpse.net/

Mark

Re: [htmltmpl] RFC: Template Tag Attributes

[Timm M. <tm...@ag...>]

At 11:30 AM 6/3/04 -0700, Mark Fuller wrote:
<>
>A good indexer would add weight to words that appear in headers/footers
>which also appear in title, meta keyword/description, headings, and
>expository content (paragraphs, list items, data terms and definitions).
>Conversely, reduce weight for words appearing in headers/footers (or any
>part of the document) that doesn't appear in outline/structural elements.
>What I really hate is all words presented to me without any weighting of
>relevance to the topic of the page (determined from its structual elements).
<>

We're pretty set on using htdig://, since we already use it for a small 
subsection of our site.  We discussed using Google site:, but (as I 
mentioned in another e-mail), Google's algorithm isn't very good for 
searching a specific site.

What I had planned on doing with Apache::QuickCMS is specify in our site 
coding standards that the template must provide certain variables for the 
content.  Specifically, title, meta-description, meta-keyword, and 
content.  Each would go in the obvious place in the HTML document.  This is 
specifically to help the search indexer do the Right Thing.

Re: [htmltmpl] RFC: Template Tag Attributes

[Mathew R. <mat...@re...>]

> >This is a common mistake that information creators think 'is a good=20
> >thing'...  The web got popular for a number of reasons - one of them =
being=20
> >"full text indexing of all content" (including headers/footers/etc).
>=20
> Why?  There is no useful information in headers/footers.  By nature of =

> using a templating system, they are the same on every page in a given=20
> section.  Including them in search results only increases the noise =
and the=20
> amount of information that needs to be indexed.

Only says you - a user of your site may find the headers and footers to =
be very useful.

> >The point is that, it is the user of the system that wants to find =
the=20
> >information - not the author telling you what you can and cant search =

> >for.   Classic example -> books used to have (and still do) an index =
in=20
> >the last couple of pages of the book, yet the user could never find =
what=20
> >they were looking for; until the book made it onto CDROM at which =
point=20
> >full-text-searching was possible.
>=20
> Almost every piece of a book is useful to search.  But what good would =
it=20
> be to search for a chapter heading?  That information is already given =
to=20
> you in the table of contents.

An index is a whole different beast to a table of contents.  And as you =
say, every peice of a book is useful to search -> thus why bother only =
indexing content that the author considers valid?  Why not just index =
_everything_, then let the user decide...

> >-> Full text searching is a _much better_ solution to search problems =
than=20
> >indexing on what YOU think is the information they want.
> >
> >Mathew
> >
> >PS.  This means, use a spider.. or even better use google via a =
"site:..."=20
> >search.
>=20
> Google PageRank is very good at searching a broad sample of sites.  =
It's=20
> not so good for individual sites.

You are kidding right?  The algorithms the Google use, already take into =
account common content on multiple pages.  OpenOffice.org use Google as =
their own site specific search engine.  As do a number of sites.  The =
only real problem with using Google is that they only spider the web =
every few weeks, thus if you update more frequently than that, you may =
have a problem.

Mathew