Re: Suggestion: ALTHREF attribute

James Green (jmkgre@essex.ac.uk)
Fri, 16 Jan 1998 12:34:07 +0000 (GMT)


From: James Green <jmkgre@essex.ac.uk>
To: www-html@w3.org
In-Reply-To: <Pine.LNX.3.96.980115171702.31698A-100000@ns.viet.net>
Message-Id: <SIMEON.9801161207.B@s1672.essex.ac.uk>
Date: Fri, 16 Jan 1998 12:34:07 +0000 (GMT)
Subject: Re: Suggestion: ALTHREF attribute


On Thu, 15 Jan 1998 18:16:23 -0800 (PST) Benjamin Franz 
<snowhare@netimages.com> wrote:

> On Thu, 15 Jan 1998, Liam Quinn wrote:
> > At 01:43 PM 15/01/98 +0000, James Green wrote:
> > >
> > ><MIRROR HREF="http://www.myname.com.au/">
> > 
> > What about
> > 
> > <LINK REL="Alternate" HREF="http://www.myname.com.au/" TYPE="text/html">
> 
> That would demand that a document be loaded *before* you could find an
> alternate to do it that way. Not very useful.
> 
> You need something that binds URLs to other URLs *before* they are
> actually fetched. And it needs to be much more flexible than LINK allows. 
> 
> Something more like:
> 
> <MIRROR SOURCEURL="http://www.mynamge.com.au/"
>         ALTURL="http://www.mnage.com/" RANDWEIGHT="0.6"
>         TYPE="text/html; q=0.7" LANG="EN; q=0.9" SIZE="6545">
> <MIRROR SOURCEURL="http://www.mynamge.com.au/"
>         ALTURL="http://www.mynamge.com.au/" RANDWEIGHT="0.8"
>         TYPE="text/html; q=0.7" LANG="EN; q=0.9" SIZE="6545">
> <MIRROR SOURCEURL="http://www.mynamge.co.jp/"
>         ALTURL="http://www.mnage.com/" RANDWEIGHT="0.5"
>         TYPE="text/html; q=0.7" LANG="JP; q=0.9" SIZE="8343">
> <MIRROR SOURCEURL="http://www.mynamge.co.jp/"
>         ALTURL="http://www.mnage.com/" RANDWEIGHT="0.4"
>         TYPE="text/xml; q=0.8" LANG="JP; q=0.9" SIZE="8343">
> 
> <MIRRORPREF SOURCEURL="http://www.mynamge.co.au/"
>         PREFERENCEORDER="LANG,DOMAIN,CONTENT-TYPE,RANDOM">
> 
> This would be an example of a resource mirrored to four locations in two
> languages and two formats, and would degrade harmlessly in older browsers.
> 
> A reference link like <A HREF="http://www.mynamge.com.au/">The Link</a>
> would be resolved in preference order first by language, 'nearest' domain
> second, Content-Type with quality ratings, with any tie being broken
> randomly finally (this would be the implict final mode if not declared
> anyhow). The alternative to RANDOM would be 'ORDERED' where the *first*
> MIRROR satisfying all the requirements would be chosen by preference.
> 
> I would love a 'MINTIME' or 'MINHOPS' option, but think that proxy caches
> would render them useless in practice.
> 
> RANDWEIGHT would allow you to bias the RANDOM option towards or away
> from particular servers to help you balance loads between servers.

Whilst this could work, it strays from usability quite wildly. The only 
place it could be used is essentially in the corporate and professional 
world. Certainly, web pages would increase in size and therefore 
downloading times (particularly sensitive issue for non-Americans who 
pay for phone usage times). An administrative nightmare it would be, 
too.

I am totally in favour of a <MIRROR> tag or the <LINK REL="ALTERNATE"> 
tags, but the deciding of what to do should be server/client 
orientated. Computers were meant as a tool to perform complex and 
repetitive things, and the web as a source of information, we should 
not complicate things unecessarily for authors like myself.

I am sure that, provided domain names are kept logical and simple (no 
.xx (xx being a country code) actually located in another country) 
browsers could decide on the closest copy to use, and based on the 
response times, could test the other nearest sites, promting the user 
for permission if such as option is set.

The elements you suggest are also prone to error and confusion should 
the author make a mistake. It would not be easy to teach, either. I 
agree that weightings could be useful, but only in the most limited of 
senses, remember most users won't be bothered too much where the 
information is gotten from, merely that it comes down reasonably 
quickly. An alternative source should the URL prove to be slow beyond a 
time perhaps allocated by the user in a browser option (defaulted to a 
reasonable amount in the http architecture) or non-existant is 
reasonable, any more would have to be decided upon by the 'intelligent' 
browser.

IMHO.



Regards,

James Green

Term e-mail: jmkgre@essex.ac.uk   |   Home e-mail: jg@cyberstorm.demon.co.uk
Homepage: http://www.cyberstorm.demon.co.uk