- From: <S.N.Brodie@ecs.soton.ac.uk>
- Date: Fri, 9 May 1997 15:44:58 +0100 (BST)
- To: www-html@w3.org
Stephanos Piperoglou wrote: > > This is where your problem is, Chad: user agents should be very lenient in > what they accept. The DTD has nothing to do with what a browser would > recognise. A good HTML viewer should be EXTREMELY fault tolerant. The whole > philosophy is being strict with what you serve and lenient with what you > accept. It is the author's responsibility to follow the rules when writing a > document. As a browser author myself, I would echo this sentiment entirely. I refer to the DTDs purely for guidance on what attributes need to be recognised and what tags need to be handled. Sadly, often this is insufficient and I have to fiddle around with Netscape and Internet Explorer - both to find out how they have chosen to represent particular constructs (to stop my users saying "It doesn't come out properly but it does in Netscape/IE") and more importantly to discover how they handle erroneous HTML. Most of the time, I don't mind the former - usually the choice of representation is fairly obvious anyway and it doesn't matter if I deviate from it. The latter is harder (eg. how do you handle pages containing illegal comments). Here, I adopted the strategy apparently followed by NS 2 - which again seemed a sensible enough strategy when I had discovered what it was. Error recovery is by far the hardest part of parsing HTML-like documents, IMHO. I'm certain that my parser would be a third of the size if it could be guaranteed to be given error free HTML to parse. -- Stewart Brodie, Electronics & Computer Science, Southampton University. http://www.ecs.soton.ac.uk/~snb94r/ http://delenn.ecs.soton.ac.uk/
Received on Friday, 9 May 1997 10:46:47 UTC