Re: LINK TYPE=override/type

Benjamin Franz (snowhare@netimages.com)
Fri, 23 Jan 1998 10:36:38 -0800 (PST)


Date: Fri, 23 Jan 1998 10:36:38 -0800 (PST)
From: Benjamin Franz <snowhare@netimages.com>
To: www-html@w3.org
In-Reply-To: <199801231803.LAA00183@underworld.bigpic.com>
Message-ID: <Pine.LNX.3.96.980123101637.8436A-100000@ns.viet.net>
Subject: Re: LINK TYPE=override/type

On Fri, 23 Jan 1998, Neil St.Laurent wrote:

> > No, it is *very* useful. It is a guide to the browser which stylesheets
> > (scripts, fonts, proprietary extensions, whatever) it can handle and
> > should load and which it should just skip because it can't do anything
> > meaningful with them anyway. Thus when a browser encounters:
> > <LINK rel=stylesheet type="text/css"   href="/stylesheets/default.css">
> > <LINK rel=stylesheet type="text/jsss"  href="/stylesheets/default.jsss">
> > <LINK rel=stylesheet type="text/xsl"   href="/stylesheets/default.xsl">
> > <LINK rel=stylesheet type="text/dsssl" href="/stylesheets/default.dsssl">
> > it *doesn't* try to load the stylesheets it can't deal with anyway..
> 
> But one of the arguments is that the HTTP header is the authoratative 
> answer for the Content-Type, this would mean that the browser would 
> have to at least get the HEAD of the above documents to verify 
> whether the TYPE is correct or not.

'A foolish consistency is the bug-a-boo of small minds.'

Its a matter *when* do you know information.

When I load the HTML document - *I don't know* the HTTP header values for
the LINKed documents. So I use the author's *HINTS* to decide which ones I
will load. This is not authoritive: It is only a HINT. A *very useful*
hint that when used properly prevents needless server accesses and makes
for a faster browsing experience.

When I load a LINKed document *then* I use the Content-Type specified by
the server. This *is* authoritive.

There is no need to be doing HEADs 'to verify'. That would defeat the
whole purpose of TYPE. *IF* the document is fetched, then the browser
will know for sure what the content-type is and then you *might* be
forced into error handling if it doesn't match the TYPE declared type.

I repeat: MISMATCHED 'TYPE' AND 'CONTENT-TYPE' IS AN ERROR.  There is no
question of 'priority' here. The situation should never arise in the first
place.

> If HTTP is authoratative then TYPE is a pointless attribute when 
> using HTTP.

No, it is far from pointless to save the browser (quite possibly a lot of)
wasted time fetching resources it couldn't use in the first place.

It is a *mis-use* of TYPE to attempt to override a server's MIME types
and it won't work anyway because you *WILL* break the most widely deployed
browsers.

-- 
Benjamin Franz