Conference 7.286::digital

The whole topic of information searching is very a vey important one.  I think 
altavista is a great tool for browsing but still has a significant way to go
when there is a need to find something IMMEDIATELY.  Consider the situation
where a customer calls up and asks a question about a product of ours.  Further
assume that altavista is used to search a database and comes up with 68423 hits!
There is no way you could ever expect someone to quickly find the answer in this
sea of information.  The flip side is if there are no hits!  Which is worse?

A third scenario is where there are a bunch of hits, but the one that contains
the information wasn't found because someone was searching for "system 
management" and the entry containing the desired information had it spelled as
"sys-mgmt".  Altavista would NEVER find it under these circumstances.

There are now more intelligent search engine that know about synonyms, stemming
of words (you could search for 'walk' and find 'walking', 'walker', 'walked',
etc). There are lots of other bells and whistles that have been been in use by
the production search engines for years as well.

My recommendation is if you need immediate access to the types of internal
databases you've mentioned, you should look at internally developed search
engines such as STARS and/or COMET, which have some or all of the features I
mentioned earlier.  You can access COMET right now at 

		errno2.alf.dec.com/cgi/v4.0

It has indexed notes files, STARS articles (symptom-solution databases for 
product support in our service centers), call handling systems (records of
customer calls/problems), news groups and a variety of other sources.  There are
conversion utilities around to allow you to take a straight ASCII database and
index it for access through these tools.  While they do NOT support the
scalability and performance of altavista, they're damn fast for the 1000's of
people who use them internally as opposed to the 1,000,000's of people that
altavista needs to support.

-mark

    re .1 
    
    but using comet you can not use anything more complex than one keyword.
    Try searching for something like loadable drivers or framework error..
    
    Mind you comet is still very fast and useful but compared to alta vista
    it leaves a little bit to be desired.
    
    /Peter
    

Let me give a totally unofficial answer ... er, guess ... since we have the
same thing in mind and I've dug a little into what's happening with Alta
Vista (about which other people certainly know more, but they seem to hang
out mostly in INTERNET_TOOLS).

Alta Vista is maintained by inspiration and hard work, and is not just a
smoothly-running package right now.  The project is under way to turn it
into a package, but that's some months away.  In the meantime, the people
who keep it running just haven't the cycles to duplicate it.

Once it's ready for field test, I sure hope we leap on it for internal use.
We'd have to be crazy not to.

Customers are already asking for it.  And we're not blind to that.

___Pete

>    but using comet you can not use anything more complex than one keyword.
>    Try searching for something like loadable drivers or framework error..

Are you sure we're talking about the same tool?  I just ran comet and found a
whole lot of entries on 'loadable drivers'.  It even found note 992 in this very
conference that contain the words 'loadable' and 'drivers'.  I'm not sure if you
can force it to treat 'loadable drivers' as a phrase, though I know in STARS
you can.  As far as complex queries I think you can do just about anything with
STARS and COMET that you can do with altavista.  Could be a good reason for
having better on-line help...

-mark    

re: .2

COMET will 'AND' any (reasonable) number of keywords.
The other boolean operators are also supported.  Please see

	http://comet.alf.dec.com/usr/comet/www/html/v4_doc.html

for COMET's actual capabilities.  

Chris
[Posted by WWW Notes gateway]

    G'day,
    
     Downunder we are the localagent for Excalibur Electronic Filing
    Software (EFS) and RetrievalWare...
    
    This is real quicjk indexing and searching.. and is tolerant of
    mis-spellings and typos since it does the searchging by pattern
    matching.
    
    The index is guaranteed to be NO MORE THAN 30% of the original data and
    the search is QUICK.
    
    We typically get less than 1 sec searches on 10,000+ pages of text on
    an Alpha 3000. onthe demo system.
    
    I heard that it searched the index to 30GB in better than 1 sec on a
    Sun platform.
    
    not cheap but it flies...
    
    try http:/www.xrs.com/
    
    
    derek

    
    RE: .5
    
    ERROR
    
    Requested document(URL
    http://comet.alf.dec.com/usr/comet/www/html/v4_doc.html) could not be
    accessed.
    
    The information server either is not accessible or is refusing to serve
    the document to you.
    

re Note 4451.3 by HERON::KAISER:

> Once it's ready for field test, I sure hope we leap on it for internal use.
> We'd have to be crazy not to.
  
        Alta Vista is certainly the way top go for internal use. 
        However, I think that if we just implemented AV as it exists
        today, i.e., indexing the Web and perhaps internal
        newsgroups, we would find that (at this time) probably the
        majority of internal information isn't yet (easily)
        web-accessible.

        Bob

    >newsgroups, we would find that (at this time) probably the
    >majority of internal information isn't yet (easily)
    >web-accessible.
    
    Is AV able to make indexes of data made availble
    through web gateways for products like ALL-IN-1 and VTX ?
    
    Would be a nice start.
    

>    http://comet.alf.dec.com/usr/comet/www/html/v4_doc.html) could not be
>    accessed.
    
I use http://errno2.alf.dec.com as a jumping off point...

-mark