RE: New Requirements Draft

On Wed, 27 Aug 1997, Yaron Goland wrote:

> Furthermore Martin, you make some very sweeping statements

Agreed, I was a bit to the sweeping side. But it has provoked
a lot of interesting reactions, it seems.


> What is a "serious" web site?

For examlpes, start with http://www.100hot.com/home.chtml


> Which of these "serious" web sites use variants?

A lot of them use language variants. Some typical US-only
sites, like usatoday, won't. But most of them will do some
other kind of configuration/version management, for example
to account for browser peculiarities and bugs :-(.
Every time Microsoft or Netscape get something wrong,
serious sites have to account for it. They can't
afford things not to work on the mainstream browsers.
And this also shows that Accept headers are not the
only thing that variants are based upon. User-Agent
is probably much more important :-(.


> What percentage
> of all web sites are "serious" and use variants?

For the 100 listed at http://www.100hot.com/home.chtml,
I would be surprised if there was any that didn't.
If you know of one, please let me know.
Not that many of them already recognize accept headers,
but as said above, that's not necessary for serving
variants.


> What commercial systems
> deploy with built in support for variant handling?

I don't know many commercial web servers. But definitely
DynaWeb or what the server from INSO is called does, and
goes to extremely great detail. Also, Apache deploys with
built in support for variant handling.


> How many people use those systems?

More and more. They haven't been around for such a long time,
and it takes time for people to realize that these things are
supported, and to learn how to use them.


> How many web sites use variants as opposed to
> having a "choose English here" tag?

See above. And the "choose English here" tag is not a definite
indication that server-side variant support is not used. Both
are necessary in a transitory phase.


> How many web sites handle primarily
> language variants by having an initial detection of the accept-language
> tag and then redirect the person to an entire site in a single language?

Quite a few. In some cases, especially for commercial sites that
target people from different language regions in different ways,
and therefore may have parts of their web sites that are
completely unrelated for two languages, this makes a lot of
sense. But there are many cases where very close parallelism
is desired. These include administrations in countries with
multiple languages (have a look e.g. at http://www.admin.ch/),
the European Commission, other supranational organizations,
but also corporations for their internal guidelines and documents.

Even for Microsoft, where a lot of marketing is very well targeted
at the local public, I expect big parts of websites (more
straightforward technical stuff) to exhibit quite a high
degree of parallelism. And with the ever-increasing streamlining
of internationalization into the overall software development
process, the reduction of local or regional versions of software,
and the world-wide coordination of launching dates, this
parallelism will increase rather than decrease.

Of course, every company or administration and their web
authoring team is free to choose whether they want to use
the mechanisms provided in HTTP, or whether they want to
create something of their own. But an authoring protocol
that forces everybody to create something of their own
where the base protocol has a well-defined mechanism
is not really something I would call a productivity
tool.

Regards,	Martin.

Received on Thursday, 28 August 1997 08:52:56 UTC