Re: why TITLE, not TITLE?

Stephanos Piperoglou (spip@hol.gr)
Sat, 10 May 1997 20:06:47 +0300 (EET DST)


Date: Sat, 10 May 1997 20:06:47 +0300 (EET DST)
From: Stephanos Piperoglou <spip@hol.gr>
To: Chad Owen Yoshikawa <chad@CS.Berkeley.EDU>
cc: papresco@calum.csclub.uwaterloo.ca, www-html@w3.org
Subject: Re: why TITLE, not TITLE?
In-Reply-To: <199705091927.MAA10934@whenever.CS.Berkeley.EDU>
Message-ID: <Pine.LNX.3.95.970510193348.308B-100000@fenchurch.hol.gr>

On Fri, 9 May 1997, Chad Owen Yoshikawa wrote:

> I'll either end up creating a 'lenient DTD' and/or a 'lenient parser'
> - I'll go off and think about it and see how hard it is to modify the
> parser itself.  I guess this is the problem of having the HTML 'browsers' come
> out before the HTML 'generators' :(

If you're making a browser, my proposal would be to avoid making your parser
DTD-based. Firstly, since you'll have to modify ANYWAY, it's a lot less work
to modify your code. Secondly, DTDs are not for browsers to use. It is the
author's responsibility to validate their HTML against a DTD, and that's why
we have DTDs, invalidating your statement about "generators". SGML tools
like SP can validate against new DTDs without any need to upgrade their
software, whilst user agents need to be updated for each new DTD. Since the
smashing majority of documents on the Web lack <!DOCTYPE... your DTD based
parser would parse according to an assumed DTD, and the assumption will
rarely be valid.

In short, the problem in your case is making a DTD based parser for a
browser. In short, don't do it. Write as liberal a code as possible for
recognising HTML.

--
Stephanos "Pippis" Piperoglou - http://users.hol.gr/~spip/index.html
I've never finished anything I began, but this time I'm