Re: character problems

Chris Lilley (Chris.Lilley@sophia.inria.fr)
Wed, 17 Jul 1996 15:42:27 +0200 (DST)


Date: Wed, 17 Jul 1996 15:42:27 +0200 (DST)
From: Chris Lilley <Chris.Lilley@sophia.inria.fr>
Message-Id: <9607171542.ZM4691@grommit.inria.fr>
In-Reply-To: Stephanos Piperoglou <stephanos@hol.gr>
To: Chris Lilley <chris@w3.org>, Stephanos Piperoglou <stephanos@hol.gr>
Subject: Re: character problems
Cc: bbos@w3.org, www-html@w3.org

On Jul 17,  3:06pm, Stephanos Piperoglou wrote:

> On Wed, 17 Jul 1996, Chris Lilley wrote:
>
> > Content-Type: text/html; charset=iso-9959-7
>
> This much I know; but the only way I know I can set HTTP headers is
> through CGI which pukes them to stdout; I don't manage the sites that my
> pages are hosted on, and how the heck can the site admin make the server
> send out the correct headers for every document?

OK, one quick way is to save all your Greek html pages with some different
extension such as .htmlg or whatever. Then declare the new MIME type which
corresponds to this extension.

An alternative for those who have access to server config is to use different
mappings from file type to MIME type for particular directories. I have seen
sites organised /foo/en/bar.html, /foo/fr/bar.html and so on. This is usually
used with the HTTP 1.1 Accept-Language HTTP header which browsers can send to
indicate the language preferences of the reader.

> Can this be accomplished
> through the META tag in the HTML header instead?


Perhaps, depends on whether the server actually alters the HTTP headers based
on the meta tags (which implies parsing each document to look for them ->
performance problem.

The W3C (was CERN) server had the concept of a meta directory. On getting a
request for /foo/bar.html it would search for /foo/.meta/bar.html which could
contain any HTTP geaders (suchg as expires, which is what I used it for). The
advantage of this approach was that if you had bar.html and bar.pdf and so on,
each one could have it's own meta data.

> How woiuld a user agent
> respond to getting HTTP headers like this AFTER it starts getting (and
> rendering) the HTML?

That would not happen, because HTTP headers are always sent before the
document.

> What do I do about it, besides contacting my provider (I also work
> for them, and I happen to know the general view point is "Make sure that
> Greek trueype fonts are installed in the control panel and change the
> encoding in the Preferences Dialog from the Options menu. Unix? Lynx?
> Mosaic? We're sorry, we've never heard of those!")

Sounds like you have some local education to do. I suggest you try the .htmlg
trick, that just requires them to add one new MIME type and the rest is under
your control.


> User agent must render my pages, written in ISO-8859-7, correctly. In
> order to do this, it must recognise thatthey are written in ISO-8859-7 and
> have the ability to render them accordingly.

Right. All this asking the user to basically guess from a menu, fiddling with
fonts until something works, is no use. However, browsers cannot really do much
else with unlabelled content, which is the major problem.

> This is done through an HTTP header
> line. I can't control what the server sends out as the document's HTTP
> header. What do I do?

You can now. Let us know how you get on.




-- 
Chris Lilley, W3C                          [ http://www.w3.org/ ]
http://www.w3.org/people/chris/                       INRIA/W3C
chris@w3.org                       2004 Rt des Lucioles / BP 93
+33 93 65 79 87            06902 Sophia Antipolis Cedex, France