Re: content negotiation by User-Agent

In article <9601291910.AA19659@aleatory.tempo.att.com>, Dave Kristo writes:
> Just some random data points....  Our www.research.att.com is
> moderately busy.  In the past 36 hours or so, we've gotten hits from
> over 40 distinct versions of Mozilla alone.  FWIW, these are the top
> five Mozilla User-Agent strings we've seen in that period (minus the
> comments), in order, with a range of hits of 3-to-1:
>     Mozilla/1.1N
>     Mozilla/2.0b6a
>     Mozilla/1.22
>     Mozilla/1.12
>     Mozilla/2.0b5
> 
> Some of the low-runners among the 40+ are version 0.9x.
> 
> One message in the "Microsoft IE -- it just gets better and better"
> thread suggested that as soon as Netscape goes to its next version,
> MS will be stuck.  I disagree.  Clearly there are a lot of old versions
> of Netscape's own browser that content providers must deal with anyway.
> Upgrades don't happen as quickly as some think.
> 

Netscape has put an expiration date in all the beta versions of their
clients since 1.0.  When that date arrives (usually timed for about
the date of a new regular release) the browser ceases to function.
Not upgrading is not an option.  Of course the "regular releases" can
be used for ever, but there aren't so many of them.  The report in
this thread was that MSIE was disguising itself as Mozilla/2.0b3.
This will expire and stop working.  After that happens only
counterfeits will use this user-agent string.

> How is User-Agent-based content negotiation supposed to work in the
> face of so many versions?
> 

One way it works by creating a file for each relevant feature, e.g. 
tables.  The file contains a list of regular expressions intended to
match browser names which can handle tables.  Then an IF/ELSE/ENDIF
syntax in the HTML document includes a table or substitute based on
whether the User-Agent matches one of the regular expressions.

Of course the files of features require maintenance, and usually
aren't well maintained.  That is the hitch.  Conditional text that is
interpreted by the browser would be better.

-- 

John Franks 	Dept of Math. Northwestern University
		john@math.nwu.edu

Received on Monday, 29 January 1996 15:35:07 UTC